compact syntax for spatial tensor ops

better IR compaction after dcp merge
remove pim.mvm op
better memory report

.gitignore

@@ -1,5 +1,17 @@
+.zed
 .idea
 **/.vscode
+
 .claude
+.codex
 AGENTS.md
+
+CMakeUserPresets.json
+
 build
+build_release
+cmake-build-debug
+cmake-build-release
+compile.sh
+
+**/__*

README.md

@@ -1,5 +1,159 @@
 # Raptor
 
+Raptor is a domain-specific MLIR compiler for neural networks (ONNX format)
+targeting in-memory computing / processing-in-memory (PIM) architectures.
+It progressively lowers ONNX-MLIR through a set of MLIR dialects down to
+target-specific artifacts (currently JSON code for the `pimsim-nn` simulator).
+
+## Overview
+
+PIM architectures perform most of the computation directly in memory.
+Raptor's first supported target is `pimsim-nn`, which simulates a chip with:
+- a shared host memory,
+- a number of cores that do most of the computation directly in their memory
+  (vector ops, vmm/mvm on ReRAM crossbars),
+- no branching instructions (branchless architecture) and no hardware loop
+  support — any repeated work (e.g. convolutions) must be unrolled into
+  explicit per-iteration instructions.
+
+Because of this, the amount of emitted instructions explodes quickly and the
+compiler must optimize aggressively at every stage to keep compilation
+tractable.
+
+A second target, `PulPim`, is planned for an accelerator with RISC-V cores
+each carrying its own in-memory computing unit and crossbars. It will live in
+a dedicated dialect (future work).
+
+### Targets and simulators
+
+`pimsim-nn` (under `backend-simulators/pim/pimsim-nn`) is used for
+**performance** estimates (latency, energy), but does not functionally execute
+the JSON code it consumes. To validate the numerical correctness of the JSON
+code produced by Raptor (or, for comparison, by the `pimcomp` compiler), we use
+a Rust simulator we maintain in-tree at
+`backend-simulators/pim/pim-simulator`.
+
+## Compilation pipeline
+
+The PIM-related sources live under `src/PIM` and the tests under `test/PIM`.
+When working on this codebase, most changes should stay confined to those
+trees (you only need to look outside, e.g. at `onnx-mlir` or `llvm`, for
+framework-level details).
+
+High-level lowering flow:
+
+```
+ONNX-MLIR ──► Spatial ──► Pim (tensor) ──► Pim (bufferized) ──► PIM JSON
+```
+
+1. **ONNX → Spatial** (`src/PIM/Conversion/ONNXToSpatial`).
+   Lowers ONNX ops into the `spat` dialect (`src/PIM/Dialect/Spatial`).
+   Spatial models a high-level spatial in-memory accelerator: vmm/mvm
+   operations are accelerated by storing a constant RHS matrix into a
+   crossbar. Crossbars cannot be re-programmed during execution, have a
+   limited fixed size, and there is a limited number of them per core.
+   Conversion patterns are split by op family under
+   `Conversion/ONNXToSpatial/Patterns/{Math,NN,Tensor}` (Conv, Gemm, MatMul,
+   Elementwise, ReduceMean, Pool, Relu, Sigmoid, Softmax, Concat, Gather,
+   Reshape, Resize, Split).
+
+2. **Spatial → Pim** (`src/PIM/Conversion/SpatialToPim`).
+   Lowers Spatial to the `pim` dialect (`src/PIM/Dialect/Pim`), which
+   materializes PIM cores (`pim.core`), inter-core communication
+   (`pim.send` / `pim.receive`), halts, and crossbar-level operations.
+
+3. **Merge compute nodes** (`src/PIM/Dialect/Spatial/Transforms/MergeComputeNodes`).
+   A DCP-inspired heuristic (Dynamic Critical Path — see the original
+   scheduling paper by Kwok & Ahmad,
+   [DCP-eScience2007](https://clouds.cis.unimelb.edu.au/papers/DCP-eScience2007.pdf))
+   that coarsens the virtual node graph and decides how to group compute
+   nodes onto cores. Our implementation is only DCP-*inspired*: it is a
+   heuristic with different assumptions from the paper (different cost
+   model, constraints from crossbar capacity / core resources, and a
+   windowed coarsening loop instead of full-graph reprioritization). The
+   `dcp-critical-window-size` option controls how many lowest-slack virtual
+   nodes each coarsening iteration considers (0 = legacy full-graph
+   analysis). Related sources: `DCPGraph/DCPAnalysis.cpp`, `Graph.cpp/.hpp`,
+   `MergeComputeNodesPass.cpp`.
+
+4. **Bufferization** (`src/PIM/Dialect/Pim/Transforms/Bufferization`).
+   Converts tensor-semantics PIM IR into memref-semantics PIM IR using the
+   standard MLIR `BufferizableOpInterface` machinery
+   (`OpBufferizationInterfaces.*`, `PimBufferization.td`).
+
+5. **PIM code generation** (`src/PIM/Pass/PimCodegen`):
+   - `HostConstantFolding` — folds host-side constants.
+   - `MaterializeHostConstantsPass` — materializes the remaining host
+     constants for emission.
+   - `VerificationPass` — checks invariants before emission.
+   - `EmitPimJsonPass` — emits the final PIM JSON consumed by `pimsim-nn`
+     and `pim-simulator`.
+
+Supporting pieces:
+- `src/PIM/Compiler` — PIM-specific compiler options (crossbar size/count,
+  core count, DCP window, experimental conv impl, concat error handling, …)
+  and `PimCodeGen` entry points.
+- `src/PIM/Common` — shared utilities (`PimCommon`, `LabeledList`).
+- `src/PIM/Pass` — auxiliary passes (`MessagePass`, `CountInstructionPass`)
+  and the `PIMPasses.h` registry used by `PimAccelerator`.
+- `src/PIM/PimAccelerator.{cpp,hpp}` — accelerator entry point: registers
+  dialects, passes, and plugs Raptor into the ONNX-MLIR driver.
+
+## Key compiler options
+
+Pass these on the `onnx-mlir` command line when compiling for PIM:
+
+- `--maccel=PIM` — select the PIM accelerator.
+- `--EmitSpatial` / `--EmitPim` / `--EmitPimBufferized` / `--EmitPimCodegen`
+  — stop the pipeline at the requested stage (default: `EmitPimCodegen`).
+- `--pim-only-codegen` — assume the input is already bufferized PIM IR and
+  run only the codegen tail.
+- `--crossbar-size=<N>` / `--crossbar-count=<N>` — crossbar dimensions and
+  per-core count.
+- `--core-count=<N>` — number of cores (`-1` picks the minimum).
+- `--dcp-critical-window-size=<N>` — DCP coarsening window (0 = legacy).
+- `--use-experimental-conv-impl` — alternative convolution lowering.
+- `--ignore-concat-error` — soft-fail corner case in `ConcatOp`.
+
+## Validation
+
+Functional validation lives in `validation/` and drives the Rust
+`pim-simulator` to compare Raptor's output against a reference.
+
+Per-operation validation (from `validation/`):
+
+```
+validate.py \
+    --raptor-path ../cmake-build-release/Release/bin/onnx-mlir \
+    --onnx-include-dir ../onnx-mlir/include
+```
+
+End-to-end network validation (example: first 4 layers of YOLOv11n):
+
+```
+validate.py \
+    --raptor-path ../cmake-build-release/Release/bin/onnx-mlir \
+    --onnx-include-dir ../onnx-mlir/include \
+    --operations-dir ./networks/yolo11n/depth_04 \
+    --crossbar-size 2048 --crossbar-count 256 --core-count 1000
+```
+
+Available networks under `validation/networks/`: `vgg16`, `yolo11n`.
+Available operations under `validation/operations/`: `add`, `conv`, `div`,
+`gather`, `gemm`, `gemv`, `mul`, `pool`, `reduce_mean`, `relu`, `resize`,
+`sigmoid`, `softmax`, `split`.
+
+## Rebuilding
+
+Release build (fast):
+
+```
+cmake --build /home/nico/raptor/raptor/cmake-build-release --target onnx-mlir -j 30
+```
+
+A slower debug build is also available — configure it the same way but with
+`-DCMAKE_BUILD_TYPE=Debug` (see installation instructions below).
+
 ## Build
 
 ### Protobuf

backend-simulators/pim/pim-simulator/Cargo.toml

@@ -13,8 +13,9 @@ name = "pimcore"
 path = "src/lib/pimcore.rs"
 
 [features]
-default = ["tracing"]
+default = []
 tracing = []
+profile_time = ["dep:plotly", "dep:comfy-table", "dep:statrs"]
 
 
 
@@ -27,3 +28,9 @@ hex = "0"
 paste = "1"
 serde = { version = "1", features = ["derive"] }
 serde_json = "1"
+statrs = {version="0.16", optional=true}
+comfy-table = {version="7.1", optional=true}
+plotly = {version="0.8", optional=true}
+rayon = "1.12.0"
+faer = "0.24.0"
+faer-traits = "0.24.0"

backend-simulators/pim/pim-simulator/src/lib/instruction_set/isa.rs

@@ -1,14 +1,19 @@
 use crate::{
-    cpu::{CPU, crossbar}, instruction_set::{
+    cpu::{CPU, crossbar},
+    instruction_set::{
         Instruction, InstructionData, InstructionStatus, InstructionType, VectorBitWith,
         helper::add_all,
-    }, memory_manager::{
+    },
+    memory_manager::{
         MemoryStorable,
         type_traits::{FromFloat, UpcastDestTraits, UpcastSlice},
-    }, tracing::TRACER, utility::{add_offset_r1, add_offset_r2, add_offset_rd}
+    },
+    tracing::TRACER,
+    utility::{add_offset_r1, add_offset_r2, add_offset_rd},
 };
 use aligned_vec::{AVec, ConstAlign};
 use anyhow::{Context, Result, ensure};
+use rayon::prelude::*;
 
 use paste::paste;
 use std::{borrow::Cow, cell::OnceCell, collections::HashMap};
@@ -76,8 +81,7 @@ pub fn functor_to_name(functor: usize) -> &'static str {
 ///////////////////////////////////////////////////////////////
 /////////////////Scalar/register Instructions//////////////////
 ///////////////////////////////////////////////////////////////
-pub fn sldi(cores: &mut CPU, data: InstructionData) -> Result<InstructionStatus> 
+pub fn sldi(cores: &mut CPU, data: InstructionData) -> Result<InstructionStatus> {
-{
     TRACER.lock().unwrap().pre_sldi(cores, data);
     let (core_indx, rd, imm) = data.get_core_rd_imm();
     let core = cores.core(core_indx);
@@ -229,25 +233,30 @@ where
     [F]: UpcastSlice<T> + UpcastSlice<M>,
     [M]: UpcastSlice<T>,
     T: UpcastDestTraits<T> + MemoryStorable,
-    M: UpcastDestTraits<M> + MemoryStorable + FromFloat,
+    // Add faer::ComplexField HERE, directly bounding M for this function only
+    M: UpcastDestTraits<M> + MemoryStorable + FromFloat + faer_traits::ComplexField,
     F: UpcastDestTraits<F> + MemoryStorable,
 {
     TRACER.lock().unwrap().pre_mvm::<F, M, T>(cores, data);
+
     let (core_indx, rd, r1, mbiw, relu, group) = data.get_core_rd_r1_mbiw_immrelu_immgroup();
     let group: usize = group.try_into().context("group can not be negative")?;
+
     let core = cores.core(core_indx);
     let r1_val = core.register(r1);
     let rd_val = core.register(rd);
+
     let (memory, crossbars) = core.get_memory_crossbar();
     let crossbar = crossbars.get_mut(group).unwrap();
     let crossbar_stored_bytes = crossbar.stored_bytes();
     let crossbar_byte_width = crossbar.width();
-    //Fix this
+
     let crossbar_elem_width = crossbar_byte_width / size_of::<M>();
     ensure!(
-        crossbar_byte_width & size_of::<M>() == 0,
+        crossbar_byte_width % size_of::<M>() == 0,
         "M not divisor of the crosbbar size"
     );
+
     let crossbar_height = crossbar.height();
     let crossbar_byte_size = crossbar_byte_width * crossbar_height;
 
@@ -257,19 +266,29 @@ where
     let load = loads[0];
     let vec: Cow<[M]> = load.up();
     let matrix = crossbar.load::<M>(crossbar_byte_size)?[0];
-    let mut res = Vec::with_capacity(crossbar_elem_width);
-    let mut partial :AVec<M, _> = AVec::<M, ConstAlign<64>>::with_capacity(64, vec.len());
-    partial.resize(vec.len(), M::from_f32(0.0));
 
-    for x in 0..crossbar_elem_width {
+    // --- FAER IMPLEMENTATION ---
-        partial[0] = vec[0] * matrix[x];
+
-        for y in 1..crossbar_height {
+    // 1. Explicitly create a Matrix Reference (MatRef)
-            partial[y] = vec[y] * matrix[y * crossbar_elem_width + x];
+    let matrix_view = faer::mat::MatRef::from_row_major_slice(
-        }
+        matrix.as_ref(),
+        crossbar_height,
+        crossbar_elem_width,
+    );
+
+    // 2. Explicitly create a Column Vector Reference (ColRef)
+    // Using `ColRef` here guarantees we don't accidentally get a RowRef (Fixes E0277)
+    let vec_view = faer::col::ColRef::from_slice(vec.as_ref());
+
+    let res_col: faer::col::Col<M> = matrix_view.transpose() * vec_view;
+
+    // 4. Convert back to standard Rust Vec
+    // try_as_slice() returns an Option<&[M]>.
+    // We can safely unwrap() because a freshly allocated, owned Col is ALWAYS contiguous!
+    let mut res: Vec<M> = (0..crossbar_elem_width).map(|i| res_col[i]).collect();
+
+    // --- END FAER ---
 
-        let mut acc = add_all(partial.as_slice());
-        res.push(acc);
-    }
     if relu != 0 {
         res.iter_mut().for_each(|x| {
             if *x < M::from_f32(0.0) {
@@ -277,12 +296,15 @@ where
             }
         });
     }
+
     ensure!(
         res.len() == crossbar_elem_width,
         "mvm generate a vector bigger thant it's requested elements"
     );
+
     let res_up: Cow<[T]> = res.as_slice().up();
     core.execute_store(rd_val, res_up.as_ref());
+
     TRACER.lock().unwrap().post_mvm::<F, M, T>(cores, data);
     Ok(InstructionStatus::Completed)
 }
@@ -533,7 +555,10 @@ where
     let r2_val = r2;
     ensure!(r2_val == 1, "Stride different than 1 not supported");
     let rd_val = core.register(rd);
-    ensure!(offset_select == 1, "Offset select cannot be different from 1");
+    ensure!(
+        offset_select == 1,
+        "Offset select cannot be different from 1"
+    );
     let r1_val = add_offset_r1(r1_val, offset_select, offset_value);
     let loads = core.reserve_load(r1_val, imm_len)?.execute_load::<F>()?;
     let load1 = loads[0];
@@ -633,7 +658,10 @@ pub fn vsoftmax(cores: &mut CPU, data: InstructionData) -> Result<InstructionSta
     panic!("You are calling a placeholder, the real call is the generic version");
 }
 
-pub(super) fn vsoftmax_impl<F, T>(cores: &mut CPU, data: InstructionData) -> Result<InstructionStatus>
+pub(super) fn vsoftmax_impl<F, T>(
+    cores: &mut CPU,
+    data: InstructionData,
+) -> Result<InstructionStatus>
 where
     [F]: UpcastSlice<T>,
     T: UpcastDestTraits<T> + MemoryStorable,
@@ -656,12 +684,11 @@ where
         .reduce(|a, b| if a > b { a } else { b })
         .unwrap();
     let exp_values: Vec<F> = load1.iter().map(|&a| (a - max_val).exp()).collect();
-    let sum = exp_values
+    let sum = exp_values.iter().copied().reduce(|a, b| a + b).unwrap();
-        .iter()
+    ensure!(
-        .copied()
+        sum > 0.0.into(),
-        .reduce(|a, b| a + b)
+        "vsoftmax normalization sum must be positive"
-        .unwrap();
+    );
-    ensure!(sum > 0.0.into(), "vsoftmax normalization sum must be positive");
     let res: Vec<F> = exp_values.iter().map(|&a| a / sum).collect();
     let res_up: Cow<[T]> = res.as_slice().up();
     core.execute_store(rd_val, res_up.as_ref());
@@ -749,12 +776,10 @@ pub fn lmv(cores: &mut CPU, data: InstructionData) -> Result<InstructionStatus>
 }
 
 pub fn send(cores: &mut CPU, data: InstructionData) -> Result<InstructionStatus> {
-    TRACER.lock().unwrap().pre_send(cores, data);
     Ok(InstructionStatus::Sending(data))
 }
 
 pub fn recv(cores: &mut CPU, data: InstructionData) -> Result<InstructionStatus> {
-    TRACER.lock().unwrap().pre_recv(cores, data);
     Ok(InstructionStatus::Reciving(data))
 }
 

backend-simulators/pim/pim-simulator/src/lib/memory_manager/type_traits.rs

@@ -55,17 +55,25 @@ pub trait HasSigm {
 
 impl HasSigm for f32 {
     fn sigm(self) -> Self {
+        if self >= 0.0 {
+            1.0 / (1.0 + (-self).exp())
+        } else {
             let ex = self.exp();
             ex / (1.0 + ex)
         }
     }
+}
 
 impl HasSigm for f64 {
     fn sigm(self) -> Self {
+        if self >= 0.0 {
+            1.0 / (1.0 + (-self).exp())
+        } else {
             let ex = self.exp();
             ex / (1.0 + ex)
         }
     }
+}
 
 pub trait HasExp {
     fn exp(self) -> Self;

backend-simulators/pim/pim-simulator/src/lib/pimcore.rs

@@ -169,6 +169,9 @@ impl<'a> Executable<'a> {
             }
         }
         print_status(cores_instructions);
+
+    #[cfg(feature = "profile_time")]
+        TRACER.lock().unwrap().report();
     }
 
     pub fn cpu(&self) -> &CPU<'a> {

backend-simulators/pim/pim-simulator/src/lib/send_recv.rs

@@ -58,6 +58,20 @@ where 'a : 'b
             && sender.internal_core == receiver.external_core
             && receiver.internal_core == sender.external_core
         {
+            {
+                let sender = &mut core_instructions[sender.internal_core];
+                let pc = sender.program_counter;
+                let inst = sender.instructions.get(pc).unwrap();
+                let data = inst.data;
+                TRACER.lock().unwrap().pre_send(cpu, data);
+            }
+            {
+                let recv = &mut core_instructions[receiver.internal_core];
+                let pc = recv.program_counter;
+                let inst = recv.instructions.get(pc).unwrap();
+                let data = inst.data;
+                TRACER.lock().unwrap().pre_recv(cpu, data);
+            }
             let [sender_core, reciver_core] =
                 cpu.get_multiple_cores([sender.internal_core, receiver.internal_core]);
             let memory = sender_core

backend-simulators/pim/pim-simulator/src/lib/tracing/disable.rs

@@ -13,7 +13,7 @@ use crate::{
 };
 use std::io::Write;
 
-#[cfg(not(feature = "tracing"))]
+#[cfg(not(any(feature = "tracing", feature = "profile_time")))]
 impl Trace {
     ///////////////////////////////////////////////////////////////
     /////////////////Scalar/register Instructions//////////////////

backend-simulators/pim/pim-simulator/src/lib/tracing/mod.rs

@@ -1,52 +1,32 @@
-mod tracing_isa;
 mod disable;
-mod pretty_print;
+#[cfg(feature = "profile_time")]
-use std::{fs::File, path::{ PathBuf}};
+mod profile;
+
+#[cfg(feature = "profile_time")]
+use profile::Trace;
+
+#[cfg(feature = "tracing")]
+mod trace;
+#[cfg(feature = "tracing")]
+use trace::Trace;
+
+use crate::Executable;
+#[cfg(not(any(feature = "tracing", feature = "profile_time")))]
+use std::path::PathBuf;
 use std::sync::{LazyLock, Mutex};
 
 
-use crate::Executable;
+#[cfg(not(any(feature = "tracing", feature = "profile_time")))]
+pub struct Trace {}
 
-#[cfg(feature = "tracing")]
+#[cfg(not(any(feature = "tracing", feature = "profile_time")))]
-pub struct Trace {
-    out_files : Vec<File>
-}
-
-
-#[cfg(feature = "tracing")]
-impl Trace {
-    fn new() -> Self {
-        Self { out_files : Vec::new()}
-    }
-
-
-    pub fn init(&mut self, num_core : usize , mut path :  PathBuf) {
-        path.pop();
-        for i in 0..num_core {
-            path.push(format!("TraceCore{}", i));
-            let file = File::create(&path).expect("Can not create file");
-            self.out_files.push(file);
-            path.pop();
-        }
-    }
-}
-
-#[cfg(not(feature = "tracing"))]
-pub struct Trace {
-}
-
-
-#[cfg(not(feature = "tracing"))]
 impl Trace {
     fn new() -> Self {
         Self {}
     }
 
-
+    pub fn init(&mut self, num_core: usize, path: PathBuf) {}
-    pub fn init(&mut self, num_core : usize, path : PathBuf ) {
-    }
 }
 
 
-pub static TRACER: LazyLock<Mutex<Trace>> = LazyLock::new(|| { Trace::new().into()});
+pub static TRACER: LazyLock<Mutex<Trace>> = LazyLock::new(|| Trace::new().into());
-

backend-simulators/pim/pim-simulator/src/lib/tracing/profile/mod.rs

@@ -0,0 +1,73 @@
+use std::{collections::HashMap, path::PathBuf, time::Instant};
+
+use crate::tracing::profile::profile_analysis::{
+    analyze_timings, generate_interactive_report, print_textual_report,
+};
+
+pub mod profile_analysis;
+pub mod profile_isa;
+
+pub struct Trace {
+    instruction_times: HashMap<String, Vec<(u128,u128)>>,
+    core_start_time: HashMap<usize, Option<Instant>>,
+    start_time: Instant,
+}
+
+impl Trace {
+    pub fn new() -> Self {
+        let mut instruction_times = HashMap::new();
+        instruction_times.insert("sldi".to_string(), Vec::with_capacity(20000));
+        instruction_times.insert("sld".to_string(), Vec::with_capacity(20000));
+        instruction_times.insert("sadd".to_string(), Vec::with_capacity(20000));
+        instruction_times.insert("ssub".to_string(), Vec::with_capacity(20000));
+        instruction_times.insert("smul".to_string(), Vec::with_capacity(20000));
+        instruction_times.insert("saddi".to_string(), Vec::with_capacity(20000));
+        instruction_times.insert("smuli".to_string(), Vec::with_capacity(20000));
+        instruction_times.insert("setbw".to_string(), Vec::with_capacity(20000));
+        instruction_times.insert("mvmul".to_string(), Vec::with_capacity(20000));
+        instruction_times.insert("vvadd".to_string(), Vec::with_capacity(20000));
+        instruction_times.insert("vvsub".to_string(), Vec::with_capacity(20000));
+        instruction_times.insert("vvmul".to_string(), Vec::with_capacity(20000));
+        instruction_times.insert("vvdmul".to_string(), Vec::with_capacity(20000));
+        instruction_times.insert("vvmax".to_string(), Vec::with_capacity(20000));
+        instruction_times.insert("vvsll".to_string(), Vec::with_capacity(20000));
+        instruction_times.insert("vvsra".to_string(), Vec::with_capacity(20000));
+        instruction_times.insert("vavg".to_string(), Vec::with_capacity(20000));
+        instruction_times.insert("vrelu".to_string(), Vec::with_capacity(20000));
+        instruction_times.insert("vtanh".to_string(), Vec::with_capacity(20000));
+        instruction_times.insert("vsigm".to_string(), Vec::with_capacity(20000));
+        instruction_times.insert("vsoftmax".to_string(), Vec::with_capacity(20000));
+        instruction_times.insert("vmv".to_string(), Vec::with_capacity(20000));
+        instruction_times.insert("vrsu".to_string(), Vec::with_capacity(20000));
+        instruction_times.insert("vrsl".to_string(), Vec::with_capacity(20000));
+        instruction_times.insert("ld".to_string(), Vec::with_capacity(20000));
+        instruction_times.insert("st".to_string(), Vec::with_capacity(20000));
+        instruction_times.insert("lldi".to_string(), Vec::with_capacity(20000));
+        instruction_times.insert("lmv".to_string(), Vec::with_capacity(20000));
+        instruction_times.insert("send".to_string(), Vec::with_capacity(20000));
+        instruction_times.insert("recv".to_string(), Vec::with_capacity(20000));
+        instruction_times.insert("wait".to_string(), Vec::with_capacity(20000));
+        instruction_times.insert("sync".to_string(), Vec::with_capacity(20000));
+        Self {
+            instruction_times,
+            core_start_time: HashMap::new(),
+            start_time: Instant::now()
+        }
+    }
+
+    pub fn init(&mut self, num_core: usize, path: PathBuf) {
+        for i in 0..num_core {
+            self.core_start_time.insert(i, None);
+        }
+    }
+
+    pub fn report(&self) {
+        let res = analyze_timings(&self.instruction_times);
+        print_textual_report(&res);
+        generate_interactive_report(
+            &self.instruction_times,
+            &["mvmul", "recv"],
+            "/tmp/report.html",
+        );
+    }
+}

backend-simulators/pim/pim-simulator/src/lib/tracing/profile/profile_analysis.rs

@@ -0,0 +1,192 @@
+use comfy_table::{Cell, Table, modifiers::UTF8_ROUND_CORNERS, presets::UTF8_FULL};
+use statrs::statistics::{Data, Distribution, Max, Min, OrderStatistics};
+use std::collections::HashMap;
+
+#[derive(Debug)]
+pub struct InstructionStats {
+    pub name: String,
+    pub count: usize,
+    pub total_time: u128,
+    pub min: f64,
+    pub max: f64,
+    pub mean: f64,
+    pub median: f64,
+    pub std_dev: f64,
+    pub cv: f64,
+    pub p95: f64,
+    pub p99: f64,
+    pub skewness: f64,
+    pub kurtosis: f64,
+}
+
+fn format_time(ns: f64) -> String {
+    if ns.is_nan() {
+        return "NaN".to_string();
+    }
+
+    if ns >= 1_000_000_000.0 {
+        format!("{:.2} s", ns / 1_000_000_000.0)
+    } else if ns >= 1_000_000.0 {
+        format!("{:.2} ms", ns / 1_000_000.0)
+    } else if ns >= 1_000.0 {
+        format!("{:.2} µs", ns / 1_000.0)
+    } else {
+        format!("{:.2} ns", ns)
+    }
+}
+
+fn calculate_skewness_kurtosis(times: &[f64], mean: f64, std_dev: f64) -> (f64, f64) {
+    let n = times.len() as f64;
+
+    if n < 4.0 || std_dev == 0.0 {
+        return (f64::NAN, f64::NAN);
+    }
+
+    let mut sum_m3 = 0.0;
+    let mut sum_m4 = 0.0;
+
+    for &x in times {
+        let deviation = x - mean;
+        sum_m3 += deviation.powi(3);
+        sum_m4 += deviation.powi(4);
+    }
+
+    let m3 = sum_m3 / n;
+    let m4 = sum_m4 / n;
+
+    let skewness = m3 / std_dev.powi(3);
+    let kurtosis = (m4 / std_dev.powi(4)) - 3.0;
+
+    (skewness, kurtosis)
+}
+
+pub fn analyze_timings(timings: &HashMap<String, Vec<(u128, u128)>>) -> Vec<InstructionStats> {
+    let mut results = Vec::new();
+
+    for (instruction, times) in timings {
+        let count = times.len();
+        if count == 0 {
+            continue;
+        }
+
+        // Extract ONLY the duration (the second element of the tuple) for stats
+        let durations: Vec<u128> = times.iter().map(|&(_, duration)| duration).collect();
+        let total_time: u128 = durations.iter().sum();
+
+        let f64_times: Vec<f64> = durations.iter().map(|&t| t as f64).collect();
+        let mut data = Data::new(f64_times.clone());
+
+        let mean = data.mean().unwrap_or(0.0);
+        let std_dev = data.std_dev().unwrap_or(0.0);
+        let cv = if mean > 0.0 { std_dev / mean } else { 0.0 };
+
+        let (skewness, kurtosis) = calculate_skewness_kurtosis(&f64_times, mean, std_dev);
+
+        results.push(InstructionStats {
+            name: instruction.clone(),
+            count,
+            total_time,
+            min: data.min(),
+            max: data.max(),
+            mean,
+            median: data.median(),
+            std_dev,
+            cv,
+            p95: data.percentile(95),
+            p99: data.percentile(99),
+            skewness,
+            kurtosis,
+        });
+    }
+
+    results.sort_by(|a, b| b.mean.partial_cmp(&a.mean).unwrap());
+    results
+}
+
+pub fn print_textual_report(stats: &[InstructionStats]) {
+    let mut table = Table::new();
+    table
+        .load_preset(UTF8_FULL)
+        .apply_modifier(UTF8_ROUND_CORNERS)
+        .set_header(vec![
+            "Instruction",
+            "Count",
+            "Total Time",
+            "Mean",
+            "Median",
+            "Min",
+            "Max",
+            "P95",
+            "P99",
+            "StdDev",
+            "CV",
+            "Skewness",
+            "Kurtosis",
+        ]);
+
+    for stat in stats {
+        table.add_row(vec![
+            Cell::new(&stat.name),
+            Cell::new(stat.count.to_string()),
+            Cell::new(format_time(stat.total_time as f64)), // Cast u128 to f64 for formatting
+            Cell::new(format_time(stat.mean)),
+            Cell::new(format_time(stat.median)),
+            Cell::new(format_time(stat.min)),
+            Cell::new(format_time(stat.max)),
+            Cell::new(format_time(stat.p95)),
+            Cell::new(format_time(stat.p99)),
+            Cell::new(format_time(stat.std_dev)),
+            Cell::new(format!("{:.3}", stat.cv)),
+            Cell::new(format!("{:.2}", stat.skewness)),
+            Cell::new(format!("{:.2}", stat.kurtosis)),
+        ]);
+    }
+
+    println!("{table}");
+}
+
+
+pub fn generate_interactive_report(
+    timings: &HashMap<String, Vec<(u128, u128)>>,
+    instructions_to_plot: &[&str], // <-- NEW: Only plot these
+    file_path: &str,
+) {
+
+use plotly::common::{Mode, Marker, Line};
+use plotly::layout::{Axis, Layout};
+use plotly::{Plot, Scatter};
+use std::collections::HashMap;
+    let mut plot = Plot::new();
+
+    for &instruction_name in instructions_to_plot {
+        // Only proceed if the instruction exists in our timings map
+        if let Some(times) = timings.get(instruction_name) {
+            let x_axis: Vec<f64> = times.iter().map(|&(ts, _)| ts as f64).collect();
+            let y_axis: Vec<f64> = times.iter().map(|&(_, dur)| dur as f64).collect();
+            
+            let text_array: Vec<String> = times.iter()
+                .map(|&(_, dur)| format_time(dur as f64))
+                .collect();
+
+            let trace = Scatter::new(x_axis, y_axis)
+                .name(instruction_name)
+                .mode(Mode::LinesMarkers)
+                .marker(Marker::new().size(4).opacity(0.6))
+                .line(Line::new().width(1.0))
+                .text_array(text_array)
+                .hover_info(plotly::common::HoverInfo::All);
+
+            plot.add_trace(trace);
+        }
+    }
+
+    let layout = Layout::new()
+        .title(plotly::common::Title::new("Simulator Timeline: Top Offenders"))
+        .x_axis(Axis::new().title(plotly::common::Title::new("Absolute Time (ns)")))
+        .y_axis(Axis::new().title(plotly::common::Title::new("Execution Duration")));
+
+    plot.set_layout(layout);
+    plot.write_html(file_path);
+    println!("🌐 Interactive timeline saved to {}", file_path);
+}
+

backend-simulators/pim/pim-simulator/src/lib/tracing/profile/profile_isa.rs

@@ -0,0 +1,364 @@
+use crate::{
+    cpu::CPU,
+    instruction_set::instruction_data::InstructionData,
+    memory_manager::{
+        MemoryStorable,
+        type_traits::{FromFloat, UpcastDestTraits, UpcastSlice},
+    },
+    tracing::Trace,
+    utility::{add_offset_r1, add_offset_rd},
+};
+use std::io::Write;
+use std::time::Instant;
+
+#[cfg(feature = "profile_time")]
+impl Trace {
+    ///////////////////////////////////////////////////////////////
+    /////////////////Scalar/register Instructions//////////////////
+    ///////////////////////////////////////////////////////////////
+
+    fn pre_impl(&mut self, cores: &mut CPU, data: InstructionData) {
+        let (core_indx, rd, imm) = data.get_core_rd_imm();
+        let core_indx = core_indx as usize;
+        if self.core_start_time.get(&core_indx).unwrap().is_none() {
+            self.core_start_time.insert(core_indx, Some(Instant::now()));
+        }
+    }
+
+    fn post_impl(&mut self, cores: &mut CPU, data: InstructionData, name: &'static str) {
+        let (core_indx, rd, imm) = data.get_core_rd_imm();
+        let core_indx = core_indx as usize;
+        let Self {
+            instruction_times,
+            core_start_time,
+            start_time,
+        } = self;
+        let now = Instant::now();
+        instruction_times
+            .get_mut(name)
+            .unwrap()
+            .push((now.duration_since(*start_time).as_nanos(), now.duration_since(core_start_time[&core_indx].unwrap()).as_nanos()));
+        self.core_start_time.insert(core_indx, None);
+    }
+
+    pub fn pre_sldi(&mut self, cores: &mut CPU, data: InstructionData) {
+        self.pre_impl(cores, data);
+    }
+
+    pub fn post_sldi(&mut self, cores: &mut CPU, data: InstructionData) {
+        self.post_impl(cores, data, "sldi");
+    }
+
+    pub fn pre_sld(&mut self, cores: &mut CPU, data: InstructionData) {
+        self.pre_impl(cores, data);
+    }
+
+    pub fn post_sld(&mut self, cores: &mut CPU, data: InstructionData) {
+        self.post_impl(cores, data, "sld");
+    }
+
+    pub fn pre_sadd(&mut self, cores: &mut CPU, data: InstructionData) {
+        self.pre_impl(cores, data);
+    }
+
+    pub fn post_sadd(&mut self, cores: &mut CPU, data: InstructionData) {
+        self.post_impl(cores, data, "sadd");
+    }
+
+    pub fn pre_ssub(&mut self, cores: &mut CPU, data: InstructionData) {
+        self.pre_impl(cores, data);
+    }
+
+    pub fn post_ssub(&mut self, cores: &mut CPU, data: InstructionData) {
+        self.post_impl(cores, data, "ssub");
+    }
+
+    pub fn pre_smul(&mut self, cores: &mut CPU, data: InstructionData) {
+        self.pre_impl(cores, data);
+    }
+
+    pub fn post_smul(&mut self, cores: &mut CPU, data: InstructionData) {
+        self.post_impl(cores, data, "smul");
+    }
+
+    pub fn pre_saddi(&mut self, cores: &mut CPU, data: InstructionData) {
+        self.pre_impl(cores, data);
+    }
+
+    pub fn post_saddi(&mut self, cores: &mut CPU, data: InstructionData) {
+        self.post_impl(cores, data, "saddi");
+    }
+
+    pub fn pre_smuli(&mut self, cores: &mut CPU, data: InstructionData) {
+        self.pre_impl(cores, data);
+    }
+
+    pub fn post_smuli(&mut self, cores: &mut CPU, data: InstructionData) {
+        self.post_impl(cores, data, "smuli");
+    }
+
+    /////////////////////////////////////////////////////////////////
+    ///////////////////Matrix/vector Instructions////////////////////
+    /////////////////////////////////////////////////////////////////
+
+    pub fn pre_setbw(&mut self, cores: &mut CPU, data: InstructionData) {
+        self.pre_impl(cores, data);
+    }
+
+    pub fn post_setbw(&mut self, cores: &mut CPU, data: InstructionData) {
+        self.post_impl(cores, data, "setbw");
+    }
+
+    pub fn pre_mvm<F, M, T>(&mut self, cores: &mut CPU, data: InstructionData)
+    where
+        [F]: UpcastSlice<T> + UpcastSlice<M>,
+        [M]: UpcastSlice<T>,
+        T: UpcastDestTraits<T> + MemoryStorable,
+        M: UpcastDestTraits<M> + MemoryStorable + FromFloat,
+        F: UpcastDestTraits<F> + MemoryStorable,
+    {
+        self.pre_impl(cores, data);
+    }
+
+    pub fn post_mvm<F, M, T>(&mut self, cores: &mut CPU, data: InstructionData)
+    where
+        [F]: UpcastSlice<T> + UpcastSlice<M>,
+        [M]: UpcastSlice<T>,
+        T: UpcastDestTraits<T> + MemoryStorable,
+        M: UpcastDestTraits<M> + MemoryStorable + FromFloat,
+        F: UpcastDestTraits<F> + MemoryStorable,
+    {
+        self.post_impl(cores, data, "mvmul");
+    }
+
+    pub fn pre_vvadd<F, T>(&mut self, cores: &mut CPU, data: InstructionData)
+    where
+        [F]: UpcastSlice<T>,
+        T: UpcastDestTraits<T> + MemoryStorable,
+        F: UpcastDestTraits<F> + MemoryStorable,
+    {
+        self.pre_impl(cores, data);
+    }
+
+    pub fn post_vvadd<F, T>(&mut self, cores: &mut CPU, data: InstructionData)
+    where
+        [F]: UpcastSlice<T>,
+        T: UpcastDestTraits<T> + MemoryStorable,
+        F: UpcastDestTraits<F> + MemoryStorable,
+    {
+        self.post_impl(cores, data, "vvadd");
+    }
+
+    pub fn pre_vvsub<F, T>(&mut self, cores: &mut CPU, data: InstructionData)
+    where
+        [F]: UpcastSlice<T>,
+        T: UpcastDestTraits<T> + MemoryStorable,
+        F: UpcastDestTraits<F> + MemoryStorable,
+    {
+        self.pre_impl(cores, data);
+    }
+
+    pub fn post_vvsub<F, T>(&mut self, cores: &mut CPU, data: InstructionData)
+    where
+        [F]: UpcastSlice<T>,
+        T: UpcastDestTraits<T> + MemoryStorable,
+        F: UpcastDestTraits<F> + MemoryStorable,
+    {
+        self.post_impl(cores, data, "vvsub");
+    }
+
+    pub fn pre_vvmul<F, T>(&mut self, cores: &mut CPU, data: InstructionData)
+    where
+        [F]: UpcastSlice<T>,
+        T: UpcastDestTraits<T> + MemoryStorable,
+        F: UpcastDestTraits<F> + MemoryStorable,
+    {
+        self.pre_impl(cores, data);
+    }
+
+    pub fn post_vvmul<F, T>(&mut self, cores: &mut CPU, data: InstructionData)
+    where
+        [F]: UpcastSlice<T>,
+        T: UpcastDestTraits<T> + MemoryStorable,
+        F: UpcastDestTraits<F> + MemoryStorable,
+    {
+        self.post_impl(cores, data, "vvmul");
+    }
+
+    pub fn pre_vvdmul<F, T>(&mut self, cores: &mut CPU, data: InstructionData)
+    where
+        [F]: UpcastSlice<T>,
+        T: UpcastDestTraits<T> + MemoryStorable,
+        F: UpcastDestTraits<F> + MemoryStorable,
+    {
+        self.pre_impl(cores, data);
+    }
+
+    pub fn post_vvdmul<F, T>(&mut self, cores: &mut CPU, data: InstructionData)
+    where
+        [F]: UpcastSlice<T>,
+        T: UpcastDestTraits<T> + MemoryStorable,
+        F: UpcastDestTraits<F> + MemoryStorable,
+    {
+        self.post_impl(cores, data, "vvdmul");
+    }
+
+    pub fn pre_vvmax<F, T>(&mut self, cores: &mut CPU, data: InstructionData)
+    where
+        [F]: UpcastSlice<T>,
+        T: UpcastDestTraits<T> + MemoryStorable,
+        F: UpcastDestTraits<F> + MemoryStorable,
+    {
+        self.pre_impl(cores, data);
+    }
+
+    pub fn post_vvmax<F, T>(&mut self, cores: &mut CPU, data: InstructionData)
+    where
+        [F]: UpcastSlice<T>,
+        T: UpcastDestTraits<T> + MemoryStorable,
+        F: UpcastDestTraits<F> + MemoryStorable,
+    {
+        self.post_impl(cores, data, "vvmax");
+    }
+
+    pub fn pre_vavg<F, T>(&mut self, cores: &mut CPU, data: InstructionData)
+    where
+        [F]: UpcastSlice<T>,
+        T: UpcastDestTraits<T> + MemoryStorable,
+        F: UpcastDestTraits<F> + MemoryStorable,
+    {
+        self.pre_impl(cores, data);
+    }
+
+    pub fn post_vavg<F, T>(&mut self, cores: &mut CPU, data: InstructionData)
+    where
+        [F]: UpcastSlice<T>,
+        T: UpcastDestTraits<T> + MemoryStorable,
+        F: UpcastDestTraits<F> + MemoryStorable,
+    {
+        self.post_impl(cores, data, "vavg");
+    }
+
+    pub fn pre_vrelu<F, T>(&mut self, cores: &mut CPU, data: InstructionData)
+    where
+        [F]: UpcastSlice<T>,
+        T: UpcastDestTraits<T> + MemoryStorable,
+        F: UpcastDestTraits<F> + MemoryStorable + From<f32>,
+    {
+        self.pre_impl(cores, data);
+    }
+
+    pub fn post_vrelu<F, T>(&mut self, cores: &mut CPU, data: InstructionData)
+    where
+        [F]: UpcastSlice<T>,
+        T: UpcastDestTraits<T> + MemoryStorable,
+        F: UpcastDestTraits<F> + MemoryStorable + From<f32>,
+    {
+        self.post_impl(cores, data, "vrelu");
+    }
+
+    pub fn pre_vtanh<F, T>(&mut self, cores: &mut CPU, data: InstructionData)
+    where
+        [F]: UpcastSlice<T>,
+        T: UpcastDestTraits<T> + MemoryStorable,
+        F: UpcastDestTraits<F> + MemoryStorable + From<f32>,
+    {
+        self.pre_impl(cores, data);
+    }
+
+    pub fn post_vtanh<F, T>(&mut self, cores: &mut CPU, data: InstructionData)
+    where
+        [F]: UpcastSlice<T>,
+        T: UpcastDestTraits<T> + MemoryStorable,
+        F: UpcastDestTraits<F> + MemoryStorable + From<f32>,
+    {
+        self.post_impl(cores, data, "vtanh");
+    }
+
+    pub fn pre_vsigm<F, T>(&mut self, cores: &mut CPU, data: InstructionData)
+    where
+        [F]: UpcastSlice<T>,
+        T: UpcastDestTraits<T> + MemoryStorable,
+        F: UpcastDestTraits<F> + MemoryStorable + From<f32>,
+    {
+        self.pre_impl(cores, data);
+    }
+
+    pub fn post_vsigm<F, T>(&mut self, cores: &mut CPU, data: InstructionData)
+    where
+        [F]: UpcastSlice<T>,
+        T: UpcastDestTraits<T> + MemoryStorable,
+        F: UpcastDestTraits<F> + MemoryStorable + From<f32>,
+    {
+        self.post_impl(cores, data, "vsigm");
+    }
+
+    pub fn pre_vsoftmax<F, T>(&mut self, cores: &mut CPU, data: InstructionData)
+    where
+        [F]: UpcastSlice<T>,
+        T: UpcastDestTraits<T> + MemoryStorable,
+        F: UpcastDestTraits<F> + MemoryStorable + From<f32>,
+    {
+        self.pre_impl(cores, data);
+    }
+
+    pub fn post_vsoftmax<F, T>(&mut self, cores: &mut CPU, data: InstructionData)
+    where
+        [F]: UpcastSlice<T>,
+        T: UpcastDestTraits<T> + MemoryStorable,
+        F: UpcastDestTraits<F> + MemoryStorable + From<f32>,
+    {
+        self.post_impl(cores, data, "vsoftmax");
+    }
+
+    /////////////////////////////////////////////////////////////////
+    /////Communication/synchronization Instructions/////////////////
+    /////////////////////////////////////////////////////////////////
+
+    pub fn pre_ld(&mut self, cores: &mut CPU, data: InstructionData) {
+        self.pre_impl(cores, data);
+    }
+    pub fn post_ld(&mut self, cores: &mut CPU, data: InstructionData) {
+        self.post_impl(cores, data, "ld");
+    }
+
+    pub fn pre_st(&mut self, cores: &mut CPU, data: InstructionData) {
+        self.pre_impl(cores, data);
+    }
+
+    pub fn post_st(&mut self, cores: &mut CPU, data: InstructionData) {
+        self.post_impl(cores, data, "st");
+    }
+
+    pub fn pre_lldi(&mut self, cores: &mut CPU, data: InstructionData) {
+        self.pre_impl(cores, data);
+    }
+
+    pub fn post_lldi(&mut self, cores: &mut CPU, data: InstructionData) {
+        self.post_impl(cores, data, "lldi");
+    }
+
+    pub fn pre_lmv(&mut self, cores: &mut CPU, data: InstructionData) {
+        self.pre_impl(cores, data);
+    }
+
+    pub fn post_lmv(&mut self, cores: &mut CPU, data: InstructionData) {
+        self.post_impl(cores, data, "lmv");
+    }
+
+    pub fn pre_send(&mut self, cores: &mut CPU, data: InstructionData) {
+        self.pre_impl(cores, data);
+    }
+
+    pub fn post_send(&mut self, cores: &mut CPU, data: InstructionData) {
+        self.post_impl(cores, data, "send");
+    }
+
+    pub fn pre_recv(&mut self, cores: &mut CPU, data: InstructionData) {
+        self.pre_impl(cores, data);
+    }
+
+    pub fn post_recv(&mut self, cores: &mut CPU, data: InstructionData) {
+        self.post_impl(cores, data, "recv");
+    }
+}

backend-simulators/pim/pim-simulator/src/lib/tracing/trace/mod.rs

@@ -0,0 +1,28 @@
+use std::{fs::File, path::PathBuf};
+
+pub mod pretty_print;
+pub mod tracing_isa;
+
+pub struct Trace {
+    out_files: Vec<File>,
+}
+
+
+impl Trace {
+    pub fn new() -> Self {
+        Self {
+            out_files: Vec::new(),
+        }
+    }
+
+    pub fn init(&mut self, num_core: usize, mut path: PathBuf) {
+        path.pop();
+        for i in 0..num_core {
+            path.push(format!("TraceCore{}", i));
+            let file = File::create(&path).expect("Can not create file");
+            self.out_files.push(file);
+            path.pop();
+        }
+    }
+}
+

backend-simulators/pim/pim-simulator/src/lib/tracing/trace/tracing_isa.rs

@@ -1,4 +1,4 @@
-use crate::tracing::pretty_print;
+use crate::{tracing::trace::pretty_print, utility::add_offset_r2};
 use std::fs::File;
 
 use crate::{
@@ -13,7 +13,6 @@ use crate::{
 };
 use std::io::Write;
 
-#[cfg(feature = "tracing")]
 impl Trace {
     ///////////////////////////////////////////////////////////////
     /////////////////Scalar/register Instructions//////////////////
@@ -284,7 +283,6 @@ impl Trace {
         M: UpcastDestTraits<M> + MemoryStorable + FromFloat,
         F: UpcastDestTraits<F> + MemoryStorable,
     {
-        use crate::tracing::pretty_print;
 
         let (core_indx, rd, r1, mbiw, relu, group) = data.get_core_rd_r1_mbiw_immrelu_immgroup();
         let file: &mut File = self
@@ -358,8 +356,6 @@ impl Trace {
         T: UpcastDestTraits<T> + MemoryStorable,
         F: UpcastDestTraits<F> + MemoryStorable,
     {
-        use crate::{tracing::pretty_print, utility::add_offset_r2};
-
         let (core_indx, rd, r1, r2, imm_len, offset_select, offset_value) =
             data.get_core_rd_r1_r2_immlen_offset();
         let file: &mut File = self
@@ -990,8 +986,6 @@ impl Trace {
     /////////////////////////////////////////////////////////////////
 
     pub fn ld_impl(&mut self, cores: &mut CPU, data: InstructionData, prefix: &'static str) {
-        use crate::tracing::pretty_print;
-
         let (core, rd, r1, _, imm_len, offset_select, offset_value) =
             data.get_core_rd_r1_r2_immlen_offset();
         let file: &mut File = self
@@ -1044,8 +1038,6 @@ impl Trace {
     }
 
     pub fn st_impl(&mut self, cores: &mut CPU, data: InstructionData, prefix: &'static str) {
-        use crate::tracing::pretty_print;
-
         let (core, rd, r1, _, imm_len, offset_select, offset_value) =
             data.get_core_rd_r1_r2_immlen_offset();
         let file: &mut File = self
@@ -1138,7 +1130,6 @@ impl Trace {
     }
 
     fn lmv_impl (&mut self, cores: &mut CPU, data: InstructionData, prefix: &'static str) {
-        use crate::tracing::pretty_print;
 
         let (core, rd, r1, _, imm_len, offset_select, offset_value) =
             data.get_core_rd_r1_r2_immlen_offset();

src/PIM/Common/CMakeLists.txt

@@ -1,5 +1,12 @@
 add_pim_library(OMPimCommon
-  PimCommon.cpp
+  IR/AddressAnalysis.cpp
+  IR/CoreBlockUtils.cpp
+  IR/EntryPointUtils.cpp
+  IR/ShapeUtils.cpp
+  IR/WeightUtils.cpp
+  Support/DebugDump.cpp
+  Support/Diagnostics.cpp
+  Support/FileSystemUtils.cpp
 
   EXCLUDE_FROM_OM_LIBS
 

src/PIM/Common/IR/AddressAnalysis.cpp

@@ -0,0 +1,259 @@
+#include "mlir/Dialect/Arith/IR/Arith.h"
+#include "mlir/Dialect/SCF/IR/SCF.h"
+#include "mlir/Interfaces/DestinationStyleOpInterface.h"
+
+#include "src/Accelerators/PIM/Common/IR/AddressAnalysis.hpp"
+#include "src/Accelerators/PIM/Common/IR/ShapeUtils.hpp"
+#include "src/Accelerators/PIM/Dialect/Pim/PimOps.hpp"
+
+namespace onnx_mlir {
+
+mlir::memref::GlobalOp lookupGlobalForGetGlobal(mlir::ModuleOp moduleOp, mlir::memref::GetGlobalOp getGlobalOp) {
+  if (!moduleOp || !getGlobalOp)
+    return {};
+  return moduleOp.lookupSymbol<mlir::memref::GlobalOp>(getGlobalOp.getName());
+}
+
+namespace {
+
+mlir::Value resolveAlias(mlir::Value value, const StaticValueKnowledge* knowledge) {
+  if (!knowledge)
+    return value;
+
+  auto iter = knowledge->aliases.find(value);
+  while (iter != knowledge->aliases.end()) {
+    value = iter->second;
+    iter = knowledge->aliases.find(value);
+  }
+  return value;
+}
+
+mlir::Value resolveLoopCarriedAliasImpl(mlir::Value value, const StaticValueKnowledge* knowledge) {
+  value = resolveAlias(value, knowledge);
+
+  if (mlir::isa<mlir::BlockArgument>(value))
+    return value;
+
+  mlir::Operation* definingOp = value.getDefiningOp();
+  if (!definingOp)
+    return value;
+
+  if (auto dpsDefiningOp = mlir::dyn_cast<mlir::DestinationStyleOpInterface>(definingOp)) {
+    if (auto result = mlir::dyn_cast<mlir::OpResult>(value))
+      if (mlir::OpOperand* tiedOperand = dpsDefiningOp.getTiedOpOperand(result))
+        return resolveLoopCarriedAliasImpl(tiedOperand->get(), knowledge);
+  }
+
+  if (auto castOp = mlir::dyn_cast<mlir::memref::CastOp>(definingOp))
+    return resolveLoopCarriedAliasImpl(castOp.getSource(), knowledge);
+  if (auto collapseOp = mlir::dyn_cast<mlir::memref::CollapseShapeOp>(definingOp))
+    return resolveLoopCarriedAliasImpl(collapseOp.getSrc(), knowledge);
+  if (auto expandOp = mlir::dyn_cast<mlir::memref::ExpandShapeOp>(definingOp))
+    return resolveLoopCarriedAliasImpl(expandOp.getSrc(), knowledge);
+
+  return value;
+}
+
+llvm::FailureOr<int64_t> resolveOpFoldResult(mlir::OpFoldResult ofr, const StaticValueKnowledge* knowledge);
+
+llvm::FailureOr<int64_t> resolveIndexValueImpl(mlir::Value value, const StaticValueKnowledge* knowledge) {
+  value = resolveAlias(value, knowledge);
+
+  if (knowledge) {
+    auto iter = knowledge->indexValues.find(value);
+    if (iter != knowledge->indexValues.end())
+      return iter->second;
+  }
+
+  auto constantOp = value.getDefiningOp<mlir::arith::ConstantOp>();
+  if (constantOp) {
+    if (auto integerAttr = mlir::dyn_cast<mlir::IntegerAttr>(constantOp.getValue()))
+      return integerAttr.getInt();
+  }
+
+  mlir::Operation* definingOp = value.getDefiningOp();
+  if (!definingOp)
+    return mlir::failure();
+
+  if (auto indexCastOp = mlir::dyn_cast<mlir::arith::IndexCastOp>(definingOp))
+    return resolveIndexValueImpl(indexCastOp.getIn(), knowledge);
+
+  if (auto addOp = mlir::dyn_cast<mlir::arith::AddIOp>(definingOp)) {
+    auto lhs = resolveIndexValueImpl(addOp.getLhs(), knowledge);
+    auto rhs = resolveIndexValueImpl(addOp.getRhs(), knowledge);
+    if (failed(lhs) || failed(rhs))
+      return mlir::failure();
+    return *lhs + *rhs;
+  }
+
+  if (auto subOp = mlir::dyn_cast<mlir::arith::SubIOp>(definingOp)) {
+    auto lhs = resolveIndexValueImpl(subOp.getLhs(), knowledge);
+    auto rhs = resolveIndexValueImpl(subOp.getRhs(), knowledge);
+    if (failed(lhs) || failed(rhs))
+      return mlir::failure();
+    return *lhs - *rhs;
+  }
+
+  if (auto mulOp = mlir::dyn_cast<mlir::arith::MulIOp>(definingOp)) {
+    auto lhs = resolveIndexValueImpl(mulOp.getLhs(), knowledge);
+    auto rhs = resolveIndexValueImpl(mulOp.getRhs(), knowledge);
+    if (failed(lhs) || failed(rhs))
+      return mlir::failure();
+    return *lhs * *rhs;
+  }
+
+  if (auto divOp = mlir::dyn_cast<mlir::arith::DivUIOp>(definingOp)) {
+    auto lhs = resolveIndexValueImpl(divOp.getLhs(), knowledge);
+    auto rhs = resolveIndexValueImpl(divOp.getRhs(), knowledge);
+    if (failed(lhs) || failed(rhs) || *rhs == 0)
+      return mlir::failure();
+    return static_cast<int64_t>(static_cast<uint64_t>(*lhs) / static_cast<uint64_t>(*rhs));
+  }
+
+  if (auto remOp = mlir::dyn_cast<mlir::arith::RemUIOp>(definingOp)) {
+    auto lhs = resolveIndexValueImpl(remOp.getLhs(), knowledge);
+    auto rhs = resolveIndexValueImpl(remOp.getRhs(), knowledge);
+    if (failed(lhs) || failed(rhs) || *rhs == 0)
+      return mlir::failure();
+    return static_cast<int64_t>(static_cast<uint64_t>(*lhs) % static_cast<uint64_t>(*rhs));
+  }
+
+  return mlir::failure();
+}
+
+llvm::FailureOr<int64_t> resolveOpFoldResult(mlir::OpFoldResult ofr, const StaticValueKnowledge* knowledge) {
+  if (auto attr = mlir::dyn_cast<mlir::Attribute>(ofr)) {
+    auto integerAttr = mlir::dyn_cast<mlir::IntegerAttr>(attr);
+    if (!integerAttr)
+      return mlir::failure();
+    return integerAttr.getInt();
+  }
+
+  return resolveIndexValueImpl(mlir::cast<mlir::Value>(ofr), knowledge);
+}
+
+llvm::FailureOr<ResolvedContiguousAddress> resolveContiguousAddressImpl(mlir::Value value,
+                                                                        const StaticValueKnowledge* knowledge) {
+  int64_t byteOffset = 0;
+  value = resolveAlias(value, knowledge);
+
+  while (true) {
+    if (mlir::isa<mlir::BlockArgument>(value))
+      return ResolvedContiguousAddress {value, byteOffset};
+
+    mlir::Operation* definingOp = value.getDefiningOp();
+    if (!definingOp)
+      return mlir::failure();
+
+    if (auto dpsDefiningOp = mlir::dyn_cast<mlir::DestinationStyleOpInterface>(definingOp)) {
+      mlir::OpOperand* tiedOperand = dpsDefiningOp.getTiedOpOperand(mlir::dyn_cast<mlir::OpResult>(value));
+      if (!tiedOperand)
+        return mlir::failure();
+      value = resolveAlias(tiedOperand->get(), knowledge);
+      continue;
+    }
+
+    if (auto forOp = mlir::dyn_cast<mlir::scf::ForOp>(definingOp)) {
+      auto result = mlir::dyn_cast<mlir::OpResult>(value);
+      if (!result)
+        return mlir::failure();
+
+      auto yieldOp = mlir::cast<mlir::scf::YieldOp>(forOp.getBody()->getTerminator());
+      mlir::Value yieldedValue = resolveLoopCarriedAliasImpl(yieldOp.getOperand(result.getResultNumber()), knowledge);
+      if (auto blockArgument = mlir::dyn_cast<mlir::BlockArgument>(yieldedValue)) {
+        if (blockArgument.getOwner() == forOp.getBody() && blockArgument.getArgNumber() > 0
+            && static_cast<unsigned>(blockArgument.getArgNumber() - 1) < forOp.getInitArgs().size()) {
+          value = resolveAlias(forOp.getInitArgs()[blockArgument.getArgNumber() - 1], knowledge);
+          continue;
+        }
+      }
+
+      value = yieldedValue;
+      continue;
+    }
+
+    if (auto subviewOp = mlir::dyn_cast<mlir::memref::SubViewOp>(definingOp)) {
+      auto sourceType = mlir::dyn_cast<mlir::MemRefType>(subviewOp.getSource().getType());
+      auto subviewType = mlir::dyn_cast<mlir::MemRefType>(subviewOp.getType());
+      if (!sourceType || !subviewType || !sourceType.hasStaticShape() || !subviewType.hasStaticShape())
+        return mlir::failure();
+
+      llvm::SmallVector<int64_t> offsets;
+      llvm::SmallVector<int64_t> sizes;
+      llvm::SmallVector<int64_t> strides;
+      offsets.reserve(subviewOp.getMixedOffsets().size());
+      sizes.reserve(subviewOp.getMixedSizes().size());
+      strides.reserve(subviewOp.getMixedStrides().size());
+
+      for (mlir::OpFoldResult offset : subviewOp.getMixedOffsets()) {
+        auto resolvedOffset = resolveOpFoldResult(offset, knowledge);
+        if (failed(resolvedOffset))
+          return mlir::failure();
+        offsets.push_back(*resolvedOffset);
+      }
+
+      for (mlir::OpFoldResult size : subviewOp.getMixedSizes()) {
+        auto resolvedSize = resolveOpFoldResult(size, knowledge);
+        if (failed(resolvedSize))
+          return mlir::failure();
+        sizes.push_back(*resolvedSize);
+      }
+
+      for (mlir::OpFoldResult stride : subviewOp.getMixedStrides()) {
+        auto resolvedStride = resolveOpFoldResult(stride, knowledge);
+        if (failed(resolvedStride))
+          return mlir::failure();
+        strides.push_back(*resolvedStride);
+      }
+
+      if (!isMemoryContiguous(sourceType.getShape(), offsets, sizes, strides))
+        return mlir::failure();
+
+      auto sourceStrides = computeRowMajorStrides(sourceType.getShape());
+      byteOffset += linearizeIndex(offsets, sourceStrides) * subviewType.getElementTypeBitWidth() / 8;
+      value = resolveAlias(subviewOp.getSource(), knowledge);
+      continue;
+    }
+
+    if (auto castOp = mlir::dyn_cast<mlir::memref::CastOp>(definingOp)) {
+      value = resolveAlias(castOp.getSource(), knowledge);
+      continue;
+    }
+    if (auto collapseOp = mlir::dyn_cast<mlir::memref::CollapseShapeOp>(definingOp)) {
+      value = resolveAlias(collapseOp.getSrc(), knowledge);
+      continue;
+    }
+    if (auto expandOp = mlir::dyn_cast<mlir::memref::ExpandShapeOp>(definingOp)) {
+      value = resolveAlias(expandOp.getSrc(), knowledge);
+      continue;
+    }
+
+    if (mlir::isa<mlir::memref::AllocOp, mlir::memref::GetGlobalOp>(definingOp))
+      return ResolvedContiguousAddress {value, byteOffset};
+
+    return mlir::failure();
+  }
+}
+
+} // namespace
+
+llvm::FailureOr<int64_t> resolveIndexValue(mlir::Value value) { return resolveIndexValueImpl(value, nullptr); }
+
+llvm::FailureOr<int64_t> resolveIndexValue(mlir::Value value, const StaticValueKnowledge& knowledge) {
+  return resolveIndexValueImpl(value, &knowledge);
+}
+
+llvm::FailureOr<ResolvedContiguousAddress> resolveContiguousAddress(mlir::Value value) {
+  return resolveContiguousAddressImpl(value, nullptr);
+}
+
+llvm::FailureOr<ResolvedContiguousAddress> resolveContiguousAddress(mlir::Value value,
+                                                                    const StaticValueKnowledge& knowledge) {
+  return resolveContiguousAddressImpl(value, &knowledge);
+}
+
+mlir::Value resolveLoopCarriedAlias(mlir::Value value, const StaticValueKnowledge& knowledge) {
+  return resolveLoopCarriedAliasImpl(value, &knowledge);
+}
+
+} // namespace onnx_mlir

src/PIM/Common/IR/AddressAnalysis.hpp

@@ -0,0 +1,43 @@
+#pragma once
+
+#include "mlir/Dialect/MemRef/IR/MemRef.h"
+#include "mlir/IR/Value.h"
+
+#include "llvm/ADT/DenseMap.h"
+
+namespace onnx_mlir {
+
+/// Describes a value as a base addressable object plus a statically known
+/// byte offset after peeling aliases, casts, and contiguous subviews.
+struct ResolvedContiguousAddress {
+  mlir::Value base;
+  int64_t byteOffset = 0;
+};
+
+/// Records compile-time facts used when interpreting address arithmetic and
+/// loop-carried aliases inside PIM regions.
+struct StaticValueKnowledge {
+  llvm::DenseMap<mlir::Value, int64_t> indexValues;
+  llvm::DenseMap<mlir::Value, mlir::Value> aliases;
+
+  StaticValueKnowledge() {}
+};
+
+mlir::memref::GlobalOp lookupGlobalForGetGlobal(mlir::ModuleOp moduleOp, mlir::memref::GetGlobalOp getGlobalOp);
+
+/// Resolves a value to contiguous backing storage when that storage can be
+/// proven statically from aliases, DPS ties, casts, and subviews.
+llvm::FailureOr<ResolvedContiguousAddress> resolveContiguousAddress(mlir::Value value);
+llvm::FailureOr<ResolvedContiguousAddress> resolveContiguousAddress(mlir::Value value,
+                                                                    const StaticValueKnowledge& knowledge);
+
+/// Statically evaluates index-like SSA values, including simple integer
+/// arithmetic and loop facts recorded in `knowledge`.
+llvm::FailureOr<int64_t> resolveIndexValue(mlir::Value value);
+llvm::FailureOr<int64_t> resolveIndexValue(mlir::Value value, const StaticValueKnowledge& knowledge);
+
+/// Follows alias, view, and DPS chains to recover the backing value of a
+/// loop-carried memref/result.
+mlir::Value resolveLoopCarriedAlias(mlir::Value value, const StaticValueKnowledge& knowledge);
+
+} // namespace onnx_mlir

src/PIM/Common/IR/CompactAsmUtils.hpp

@@ -0,0 +1,745 @@
+#ifndef ONNX_MLIR_PIM_COMPACT_ASM_UTILS_HPP
+#define ONNX_MLIR_PIM_COMPACT_ASM_UTILS_HPP
+
+#include "mlir/IR/OpImplementation.h"
+#include "mlir/IR/Value.h"
+#include "mlir/Support/LLVM.h"
+
+#include "llvm/ADT/ArrayRef.h"
+#include "llvm/ADT/STLExtras.h"
+#include "llvm/ADT/SmallVector.h"
+#include "llvm/ADT/StringRef.h"
+#include "llvm/Support/LogicalResult.h"
+
+namespace onnx_mlir {
+namespace compact_asm {
+
+using namespace mlir;
+
+enum class ListDelimiter {
+  Square,
+  Paren
+};
+
+inline ParseResult parseOpenDelimiter(OpAsmParser& parser, ListDelimiter delimiter) {
+  if (delimiter == ListDelimiter::Square)
+    return parser.parseLSquare();
+  return parser.parseLParen();
+}
+
+inline ParseResult parseOptionalCloseDelimiter(OpAsmParser& parser, ListDelimiter delimiter) {
+  if (delimiter == ListDelimiter::Square)
+    return parser.parseOptionalRSquare();
+  return parser.parseOptionalRParen();
+}
+
+template <typename StreamT>
+inline void printOpenDelimiter(StreamT& stream, ListDelimiter delimiter) {
+  stream << (delimiter == ListDelimiter::Square ? "[" : "(");
+}
+
+template <typename StreamT>
+inline void printCloseDelimiter(StreamT& stream, ListDelimiter delimiter) {
+  stream << (delimiter == ListDelimiter::Square ? "]" : ")");
+}
+
+template <typename EntryT, typename ParseEntryFn>
+inline ParseResult parseCompressedRepeatedList(OpAsmParser& parser,
+                                               ListDelimiter delimiter,
+                                               SmallVectorImpl<EntryT>& entries,
+                                               ParseEntryFn parseEntry) {
+  if (parseOpenDelimiter(parser, delimiter))
+    return failure();
+  if (succeeded(parseOptionalCloseDelimiter(parser, delimiter)))
+    return success();
+
+  while (true) {
+    EntryT entry;
+    if (parseEntry(entry))
+      return failure();
+
+    int64_t repeatCount = 1;
+    if (succeeded(parser.parseOptionalKeyword("x"))) {
+      if (parser.parseInteger(repeatCount) || repeatCount <= 0)
+        return parser.emitError(parser.getCurrentLocation(), "repeat count after 'x' must be positive");
+    }
+    for (int64_t index = 0; index < repeatCount; ++index)
+      entries.push_back(entry);
+
+    if (succeeded(parseOptionalCloseDelimiter(parser, delimiter)))
+      return success();
+    if (parser.parseComma())
+      return failure();
+  }
+}
+
+template <typename IntT>
+inline ParseResult
+parseCompressedIntegerEntries(OpAsmParser& parser, ListDelimiter delimiter, SmallVectorImpl<IntT>& values) {
+  if (succeeded(parseOptionalCloseDelimiter(parser, delimiter)))
+    return success();
+
+  while (true) {
+    if (succeeded(parser.parseOptionalLParen())) {
+      SmallVector<IntT> subgroup;
+      if (parseCompressedIntegerEntries(parser, ListDelimiter::Paren, subgroup))
+        return failure();
+
+      int64_t repeatCount = 1;
+      if (succeeded(parser.parseOptionalKeyword("x"))) {
+        if (parser.parseInteger(repeatCount) || repeatCount <= 0)
+          return parser.emitError(parser.getCurrentLocation(), "repeat count after 'x' must be positive");
+      }
+      for (int64_t repeat = 0; repeat < repeatCount; ++repeat)
+        llvm::append_range(values, subgroup);
+    }
+    else {
+      int64_t first = 0;
+      if (parser.parseInteger(first))
+        return failure();
+
+      if (succeeded(parser.parseOptionalKeyword("to"))) {
+        int64_t last = 0;
+        if (parser.parseInteger(last) || last < first)
+          return parser.emitError(parser.getCurrentLocation(), "invalid ascending range");
+
+        int64_t step = 1;
+        if (succeeded(parser.parseOptionalKeyword("by"))) {
+          if (parser.parseInteger(step) || step <= 0)
+            return parser.emitError(parser.getCurrentLocation(), "step after 'by' must be positive");
+        }
+        int64_t repeatCount = 1;
+        if (succeeded(parser.parseOptionalKeyword("x"))) {
+          if (parser.parseInteger(repeatCount) || repeatCount <= 0)
+            return parser.emitError(parser.getCurrentLocation(), "repeat count after 'x' must be positive");
+        }
+        if ((last - first) % step != 0) {
+          return parser.emitError(parser.getCurrentLocation(),
+                                  "range end must be reachable from start using the given step");
+        }
+
+        for (int64_t value = first; value <= last; value += step)
+          for (int64_t index = 0; index < repeatCount; ++index)
+            values.push_back(static_cast<IntT>(value));
+      }
+      else {
+        int64_t repeatCount = 1;
+        if (succeeded(parser.parseOptionalKeyword("x"))) {
+          if (parser.parseInteger(repeatCount) || repeatCount <= 0)
+            return parser.emitError(parser.getCurrentLocation(), "repeat count after 'x' must be positive");
+        }
+        for (int64_t index = 0; index < repeatCount; ++index)
+          values.push_back(static_cast<IntT>(first));
+      }
+    }
+
+    if (succeeded(parseOptionalCloseDelimiter(parser, delimiter)))
+      return success();
+    if (parser.parseComma())
+      return failure();
+  }
+}
+
+template <typename IntT>
+inline ParseResult
+parseCompressedIntegerSequence(OpAsmParser& parser, ListDelimiter delimiter, SmallVectorImpl<IntT>& values) {
+  if (parseOpenDelimiter(parser, delimiter))
+    return failure();
+  return parseCompressedIntegerEntries(parser, delimiter, values);
+}
+
+template <typename RangeT, typename PrintEntryFn>
+inline void printCompressedEqualRuns(OpAsmPrinter& printer, RangeT entries, PrintEntryFn printEntry) {
+  for (size_t index = 0; index < entries.size();) {
+    size_t runEnd = index + 1;
+    while (runEnd < entries.size() && entries[runEnd] == entries[index])
+      ++runEnd;
+
+    if (index != 0)
+      printer << ", ";
+    printEntry(entries[index]);
+    size_t runLength = runEnd - index;
+    if (runLength > 1)
+      printer << " x" << runLength;
+    index = runEnd;
+  }
+}
+
+template <typename StreamT, typename IntT>
+inline void printCompressedIntegerEntries(StreamT& stream, ArrayRef<IntT> values) {
+  struct FlatCompression {
+    enum class Kind {
+      Single,
+      EqualRun,
+      Progression
+    };
+
+    Kind kind = Kind::Single;
+    size_t covered = 1;
+    size_t repeatCount = 1;
+    size_t progressionValueCount = 1;
+    int64_t step = 1;
+    IntT firstValue {};
+    IntT lastValue {};
+  };
+
+  auto computeFlatCompression = [&](size_t start) {
+    FlatCompression compression;
+    compression.firstValue = values[start];
+    compression.lastValue = values[start];
+
+    auto findEqualRunEnd = [&](size_t runStart) {
+      size_t runEnd = runStart + 1;
+      while (runEnd < values.size() && values[runEnd] == values[runStart])
+        ++runEnd;
+      return runEnd;
+    };
+
+    size_t firstRunEnd = findEqualRunEnd(start);
+    compression.repeatCount = firstRunEnd - start;
+    size_t progressionEnd = firstRunEnd;
+    int64_t step = 0;
+    IntT lastValue = values[start];
+
+    if (firstRunEnd < values.size()) {
+      size_t secondRunEnd = findEqualRunEnd(firstRunEnd);
+      step = static_cast<int64_t>(values[firstRunEnd]) - static_cast<int64_t>(values[start]);
+      if (step > 0 && secondRunEnd - firstRunEnd == compression.repeatCount) {
+        progressionEnd = secondRunEnd;
+        lastValue = values[firstRunEnd];
+        size_t currentRunStart = secondRunEnd;
+        while (currentRunStart < values.size()) {
+          size_t currentRunEnd = findEqualRunEnd(currentRunStart);
+          if (currentRunEnd - currentRunStart != compression.repeatCount)
+            break;
+          if (static_cast<int64_t>(values[currentRunStart]) != static_cast<int64_t>(lastValue) + step)
+            break;
+          lastValue = values[currentRunStart];
+          progressionEnd = currentRunEnd;
+          currentRunStart = currentRunEnd;
+        }
+      }
+      else {
+        step = 0;
+      }
+    }
+
+    compression.covered = 1;
+    if (progressionEnd > firstRunEnd) {
+      size_t progressionValueCount = (progressionEnd - start) / compression.repeatCount;
+      if (progressionValueCount >= 3) {
+        compression.kind = FlatCompression::Kind::Progression;
+        compression.covered = progressionEnd - start;
+        compression.progressionValueCount = progressionValueCount;
+        compression.step = step;
+        compression.lastValue = lastValue;
+        return compression;
+      }
+    }
+
+    if (compression.repeatCount > 1) {
+      compression.kind = FlatCompression::Kind::EqualRun;
+      compression.covered = compression.repeatCount;
+      return compression;
+    }
+
+    return compression;
+  };
+
+  auto findRepeatedSublist = [&](size_t start) {
+    size_t bestLength = 0;
+    size_t bestRepeatCount = 1;
+    size_t remaining = values.size() - start;
+
+    for (size_t length = 2; length * 2 <= remaining; ++length) {
+      size_t repeatCount = 1;
+      ArrayRef<IntT> candidate = values.slice(start, length);
+      while (start + (repeatCount + 1) * length <= values.size()
+             && llvm::equal(candidate, values.slice(start + repeatCount * length, length))) {
+        ++repeatCount;
+      }
+
+      if (repeatCount <= 1)
+        continue;
+
+      size_t covered = length * repeatCount;
+      size_t bestCovered = bestLength * bestRepeatCount;
+      if (covered > bestCovered || (covered == bestCovered && length < bestLength)) {
+        bestLength = length;
+        bestRepeatCount = repeatCount;
+      }
+    }
+
+    return std::pair(bestLength, bestRepeatCount);
+  };
+
+  for (size_t index = 0; index < values.size();) {
+    if (index != 0)
+      stream << ", ";
+
+    FlatCompression flat = computeFlatCompression(index);
+    auto [sublistLength, sublistRepeatCount] = findRepeatedSublist(index);
+    size_t repeatedSublistCoverage = sublistLength * sublistRepeatCount;
+    if (sublistRepeatCount > 1 && sublistLength > 1 && repeatedSublistCoverage > flat.covered) {
+      printOpenDelimiter(stream, ListDelimiter::Paren);
+      printCompressedIntegerEntries(stream, values.slice(index, sublistLength));
+      printCloseDelimiter(stream, ListDelimiter::Paren);
+      stream << " x" << sublistRepeatCount;
+      index += repeatedSublistCoverage;
+      continue;
+    }
+
+    switch (flat.kind) {
+    case FlatCompression::Kind::Progression:
+      stream << flat.firstValue << " to " << flat.lastValue;
+      if (flat.step != 1)
+        stream << " by " << flat.step;
+      if (flat.repeatCount > 1)
+        stream << " x" << flat.repeatCount;
+      index += flat.covered;
+      break;
+    case FlatCompression::Kind::EqualRun:
+      stream << flat.firstValue << " x" << flat.repeatCount;
+      index += flat.covered;
+      break;
+    case FlatCompression::Kind::Single:
+      stream << flat.firstValue;
+      index += flat.covered;
+      break;
+    }
+  }
+}
+
+template <typename StreamT, typename IntT>
+inline void printCompressedIntegerSequence(StreamT& stream, ArrayRef<IntT> values, ListDelimiter delimiter) {
+  printOpenDelimiter(stream, delimiter);
+  printCompressedIntegerEntries(stream, values);
+  printCloseDelimiter(stream, delimiter);
+}
+
+template <typename IntT>
+inline ParseResult parseCompressedIntegerList(OpAsmParser& parser, SmallVectorImpl<IntT>& values) {
+  return parseCompressedIntegerSequence(parser, ListDelimiter::Square, values);
+}
+
+template <typename IntT>
+inline void printCompressedIntegerList(OpAsmPrinter& printer, ArrayRef<IntT> values) {
+  printCompressedIntegerSequence(printer, values, ListDelimiter::Square);
+}
+
+inline void printCompressedValueSequence(OpAsmPrinter& printer, ValueRange values) {
+  for (size_t index = 0; index < values.size();) {
+    size_t equalRunEnd = index + 1;
+    while (equalRunEnd < values.size() && values[equalRunEnd] == values[index])
+      ++equalRunEnd;
+
+    if (index != 0)
+      printer << ", ";
+    if (equalRunEnd - index > 1) {
+      printer.printOperand(values[index]);
+      printer << " x" << (equalRunEnd - index);
+      index = equalRunEnd;
+      continue;
+    }
+
+    size_t rangeEnd = index + 1;
+    if (auto firstResult = dyn_cast<OpResult>(values[index])) {
+      while (rangeEnd < values.size()) {
+        auto nextResult = dyn_cast<OpResult>(values[rangeEnd]);
+        if (!nextResult || nextResult.getOwner() != firstResult.getOwner()
+            || nextResult.getResultNumber() != firstResult.getResultNumber() + (rangeEnd - index)) {
+          break;
+        }
+        ++rangeEnd;
+      }
+    }
+    else if (auto firstArg = dyn_cast<BlockArgument>(values[index])) {
+      while (rangeEnd < values.size()) {
+        auto nextArg = dyn_cast<BlockArgument>(values[rangeEnd]);
+        if (!nextArg || nextArg.getOwner() != firstArg.getOwner()
+            || nextArg.getArgNumber() != firstArg.getArgNumber() + (rangeEnd - index)) {
+          break;
+        }
+        ++rangeEnd;
+      }
+    }
+
+    printer.printOperand(values[index]);
+    if (rangeEnd - index >= 3) {
+      printer << " to ";
+      printer.printOperand(values[rangeEnd - 1]);
+    }
+    else if (rangeEnd - index == 2) {
+      printer << ", ";
+      printer.printOperand(values[index + 1]);
+    }
+    index = rangeEnd;
+  }
+}
+
+inline void printCompressedTypeSequence(OpAsmPrinter& printer, TypeRange types) {
+  printCompressedEqualRuns(printer, types, [&](Type type) { printer.printType(type); });
+}
+
+inline void printCompressedValueList(OpAsmPrinter& printer, ValueRange values, ListDelimiter delimiter) {
+  printOpenDelimiter(printer, delimiter);
+  printCompressedValueSequence(printer, values);
+  printCloseDelimiter(printer, delimiter);
+}
+
+inline void printCompressedTypeList(OpAsmPrinter& printer, TypeRange types, ListDelimiter delimiter) {
+  printOpenDelimiter(printer, delimiter);
+  printCompressedTypeSequence(printer, types);
+  printCloseDelimiter(printer, delimiter);
+}
+
+inline ParseResult parseCompressedTypeSequence(OpAsmParser& parser, SmallVectorImpl<Type>& types, bool allowEmpty) {
+  Type firstType;
+  OptionalParseResult firstTypeResult = parser.parseOptionalType(firstType);
+  if (!firstTypeResult.has_value()) {
+    if (allowEmpty)
+      return success();
+    return parser.emitError(parser.getCurrentLocation(), "expected type");
+  }
+  if (failed(*firstTypeResult))
+    return failure();
+
+  auto appendType = [&](Type type) -> ParseResult {
+    int64_t repeatCount = 1;
+    if (succeeded(parser.parseOptionalKeyword("x"))) {
+      if (parser.parseInteger(repeatCount) || repeatCount <= 0)
+        return parser.emitError(parser.getCurrentLocation(), "repeat count after 'x' must be positive");
+    }
+    for (int64_t index = 0; index < repeatCount; ++index)
+      types.push_back(type);
+    return success();
+  };
+
+  if (appendType(firstType))
+    return failure();
+  while (succeeded(parser.parseOptionalComma())) {
+    Type nextType;
+    if (parser.parseType(nextType) || appendType(nextType))
+      return failure();
+  }
+  return success();
+}
+
+inline ParseResult parseCompressedOperandEntryWithFirst(OpAsmParser& parser,
+                                                        OpAsmParser::UnresolvedOperand firstOperand,
+                                                        SmallVectorImpl<OpAsmParser::UnresolvedOperand>& operands) {
+  if (succeeded(parser.parseOptionalKeyword("to"))) {
+    OpAsmParser::UnresolvedOperand lastOperand;
+    if (parser.parseOperand(lastOperand))
+      return failure();
+    if (firstOperand.name != lastOperand.name || firstOperand.number > lastOperand.number)
+      return parser.emitError(parser.getCurrentLocation(), "invalid operand range");
+    for (unsigned number = firstOperand.number; number <= lastOperand.number; ++number)
+      operands.push_back({firstOperand.location, firstOperand.name, number});
+    return success();
+  }
+
+  int64_t repeatCount = 1;
+  if (succeeded(parser.parseOptionalKeyword("x"))) {
+    if (parser.parseInteger(repeatCount) || repeatCount <= 0)
+      return parser.emitError(parser.getCurrentLocation(), "repeat count after 'x' must be positive");
+  }
+  for (int64_t index = 0; index < repeatCount; ++index)
+    operands.push_back(firstOperand);
+  return success();
+}
+
+inline ParseResult parseOneCompressedOperandEntry(OpAsmParser& parser,
+                                                  SmallVectorImpl<OpAsmParser::UnresolvedOperand>& operands) {
+  OpAsmParser::UnresolvedOperand firstOperand;
+  if (parser.parseOperand(firstOperand))
+    return failure();
+  return parseCompressedOperandEntryWithFirst(parser, firstOperand, operands);
+}
+
+inline ParseResult parseCompressedOperandList(OpAsmParser& parser,
+                                              ListDelimiter delimiter,
+                                              SmallVectorImpl<OpAsmParser::UnresolvedOperand>& operands) {
+  if (parseOpenDelimiter(parser, delimiter))
+    return failure();
+  if (succeeded(parseOptionalCloseDelimiter(parser, delimiter)))
+    return success();
+
+  while (true) {
+    if (parseOneCompressedOperandEntry(parser, operands))
+      return failure();
+    if (succeeded(parseOptionalCloseDelimiter(parser, delimiter)))
+      return success();
+    if (parser.parseComma())
+      return failure();
+  }
+}
+
+inline ParseResult parseCompressedOperandSequence(OpAsmParser& parser,
+                                                  SmallVectorImpl<OpAsmParser::UnresolvedOperand>& operands) {
+  if (parseOneCompressedOperandEntry(parser, operands))
+    return failure();
+  while (succeeded(parser.parseOptionalComma()))
+    if (parseOneCompressedOperandEntry(parser, operands))
+      return failure();
+  return success();
+}
+
+inline ParseResult parseCompressedTypeList(OpAsmParser& parser, ListDelimiter delimiter, SmallVectorImpl<Type>& types) {
+  if (parseOpenDelimiter(parser, delimiter))
+    return failure();
+  if (succeeded(parseOptionalCloseDelimiter(parser, delimiter)))
+    return success();
+
+  if (parseCompressedTypeSequence(parser, types, /*allowEmpty=*/false))
+    return failure();
+  return parseOptionalCloseDelimiter(parser, delimiter);
+}
+
+inline bool hasRepeatedTuple(ValueRange values, size_t tupleSize) {
+  if (tupleSize == 0 || values.empty() || values.size() % tupleSize != 0)
+    return false;
+
+  SmallVector<Value> valueVec(values.begin(), values.end());
+  ArrayRef<Value> tuple(valueVec.data(), tupleSize);
+  for (size_t index = tupleSize; index < values.size(); index += tupleSize)
+    if (!llvm::equal(tuple, ArrayRef<Value>(valueVec).slice(index, tupleSize)))
+      return false;
+  return true;
+}
+
+inline bool hasRepeatedTuple(TypeRange types, size_t tupleSize) {
+  if (tupleSize == 0 || types.empty() || types.size() % tupleSize != 0)
+    return false;
+
+  SmallVector<Type> typeVec(types.begin(), types.end());
+  ArrayRef<Type> tuple(typeVec.data(), tupleSize);
+  for (size_t index = tupleSize; index < types.size(); index += tupleSize)
+    if (!llvm::equal(tuple, ArrayRef<Type>(typeVec).slice(index, tupleSize)))
+      return false;
+  return true;
+}
+
+inline void printValueTupleRun(OpAsmPrinter& printer, ValueRange values, size_t tupleSize, ListDelimiter delimiter) {
+  printOpenDelimiter(printer, delimiter);
+  printOpenDelimiter(printer, ListDelimiter::Paren);
+  for (size_t index = 0; index < tupleSize; ++index) {
+    if (index != 0)
+      printer << ", ";
+    printer.printOperand(values[index]);
+  }
+  printCloseDelimiter(printer, ListDelimiter::Paren);
+  printer << " x" << (values.size() / tupleSize);
+  printCloseDelimiter(printer, delimiter);
+}
+
+inline void printTypeTupleRun(OpAsmPrinter& printer, TypeRange types, size_t tupleSize, ListDelimiter delimiter) {
+  printOpenDelimiter(printer, delimiter);
+  printOpenDelimiter(printer, ListDelimiter::Paren);
+  for (size_t index = 0; index < tupleSize; ++index) {
+    if (index != 0)
+      printer << ", ";
+    printer.printType(types[index]);
+  }
+  printCloseDelimiter(printer, ListDelimiter::Paren);
+  printer << " x" << (types.size() / tupleSize);
+  printCloseDelimiter(printer, delimiter);
+}
+
+inline ParseResult parseCompressedOrTupleOperandList(OpAsmParser& parser,
+                                                     ListDelimiter delimiter,
+                                                     SmallVectorImpl<OpAsmParser::UnresolvedOperand>& operands) {
+  if (parseOpenDelimiter(parser, delimiter))
+    return failure();
+  if (succeeded(parseOptionalCloseDelimiter(parser, delimiter)))
+    return success();
+
+  if (succeeded(parser.parseOptionalLParen())) {
+    SmallVector<OpAsmParser::UnresolvedOperand> tupleOperands;
+    if (parseCompressedOperandSequence(parser, tupleOperands) || parser.parseRParen())
+      return failure();
+
+    int64_t repeatCount = 1;
+    if (succeeded(parser.parseOptionalKeyword("x"))) {
+      if (parser.parseInteger(repeatCount) || repeatCount <= 0)
+        return parser.emitError(parser.getCurrentLocation(), "repeat count after 'x' must be positive");
+    }
+    for (int64_t repeat = 0; repeat < repeatCount; ++repeat)
+      llvm::append_range(operands, tupleOperands);
+
+    while (succeeded(parser.parseOptionalComma())) {
+      if (parser.parseLParen())
+        return failure();
+      tupleOperands.clear();
+      if (parseCompressedOperandSequence(parser, tupleOperands) || parser.parseRParen())
+        return failure();
+
+      repeatCount = 1;
+      if (succeeded(parser.parseOptionalKeyword("x"))) {
+        if (parser.parseInteger(repeatCount) || repeatCount <= 0)
+          return parser.emitError(parser.getCurrentLocation(), "repeat count after 'x' must be positive");
+      }
+      for (int64_t repeat = 0; repeat < repeatCount; ++repeat)
+        llvm::append_range(operands, tupleOperands);
+    }
+    return parseOptionalCloseDelimiter(parser, delimiter);
+  }
+
+  while (true) {
+    if (parseOneCompressedOperandEntry(parser, operands))
+      return failure();
+    if (succeeded(parseOptionalCloseDelimiter(parser, delimiter)))
+      return success();
+    if (parser.parseComma())
+      return failure();
+  }
+}
+
+inline ParseResult
+parseCompressedOrTupleTypeList(OpAsmParser& parser, ListDelimiter delimiter, SmallVectorImpl<Type>& types) {
+  if (parseOpenDelimiter(parser, delimiter))
+    return failure();
+  if (succeeded(parseOptionalCloseDelimiter(parser, delimiter)))
+    return success();
+
+  if (succeeded(parser.parseOptionalLParen())) {
+    SmallVector<Type> tupleTypes;
+    if (parseCompressedTypeSequence(parser, tupleTypes, /*allowEmpty=*/false) || parser.parseRParen())
+      return failure();
+
+    int64_t repeatCount = 1;
+    if (succeeded(parser.parseOptionalKeyword("x"))) {
+      if (parser.parseInteger(repeatCount) || repeatCount <= 0)
+        return parser.emitError(parser.getCurrentLocation(), "repeat count after 'x' must be positive");
+    }
+    for (int64_t repeat = 0; repeat < repeatCount; ++repeat)
+      llvm::append_range(types, tupleTypes);
+
+    while (succeeded(parser.parseOptionalComma())) {
+      if (parser.parseLParen())
+        return failure();
+      tupleTypes.clear();
+      if (parseCompressedTypeSequence(parser, tupleTypes, /*allowEmpty=*/false) || parser.parseRParen())
+        return failure();
+
+      repeatCount = 1;
+      if (succeeded(parser.parseOptionalKeyword("x"))) {
+        if (parser.parseInteger(repeatCount) || repeatCount <= 0)
+          return parser.emitError(parser.getCurrentLocation(), "repeat count after 'x' must be positive");
+      }
+      for (int64_t repeat = 0; repeat < repeatCount; ++repeat)
+        llvm::append_range(types, tupleTypes);
+    }
+    return parseOptionalCloseDelimiter(parser, delimiter);
+  }
+
+  while (true) {
+    Type type;
+    if (parser.parseType(type))
+      return failure();
+
+    int64_t repeatCount = 1;
+    if (succeeded(parser.parseOptionalKeyword("x"))) {
+      if (parser.parseInteger(repeatCount) || repeatCount <= 0)
+        return parser.emitError(parser.getCurrentLocation(), "repeat count after 'x' must be positive");
+    }
+    for (int64_t repeat = 0; repeat < repeatCount; ++repeat)
+      types.push_back(type);
+
+    if (succeeded(parseOptionalCloseDelimiter(parser, delimiter)))
+      return success();
+    if (parser.parseComma())
+      return failure();
+  }
+}
+
+inline void printArgumentBindings(OpAsmPrinter& printer, Block& block, ValueRange operands) {
+  if (block.getNumArguments() == 0) {
+    printer << "() = ()";
+    return;
+  }
+
+  if (block.getNumArguments() == 1) {
+    printer.printOperand(block.getArgument(0));
+    printer << " = ";
+    printCompressedValueList(printer, operands, ListDelimiter::Paren);
+    return;
+  }
+
+  printCompressedValueList(printer, ValueRange(block.getArguments()), ListDelimiter::Paren);
+  printer << " = ";
+  printCompressedValueList(printer, operands, ListDelimiter::Paren);
+}
+
+inline ParseResult parseCompressedArgumentEntryWithFirst(OpAsmParser& parser,
+                                                         OpAsmParser::Argument firstArgument,
+                                                         SmallVectorImpl<OpAsmParser::Argument>& arguments) {
+  if (succeeded(parser.parseOptionalKeyword("to"))) {
+    OpAsmParser::Argument lastArgument;
+    if (parser.parseArgument(lastArgument))
+      return failure();
+    if (firstArgument.ssaName.name != lastArgument.ssaName.name
+        || firstArgument.ssaName.number > lastArgument.ssaName.number) {
+      return parser.emitError(parser.getCurrentLocation(), "invalid argument range");
+    }
+    for (unsigned number = firstArgument.ssaName.number; number <= lastArgument.ssaName.number; ++number) {
+      OpAsmParser::Argument argument;
+      argument.ssaName = {firstArgument.ssaName.location, firstArgument.ssaName.name, number};
+      arguments.push_back(argument);
+    }
+    return success();
+  }
+
+  arguments.push_back(firstArgument);
+  return success();
+}
+
+inline ParseResult parseOneCompressedArgumentEntry(OpAsmParser& parser,
+                                                   SmallVectorImpl<OpAsmParser::Argument>& arguments) {
+  OpAsmParser::Argument firstArgument;
+  if (parser.parseArgument(firstArgument))
+    return failure();
+  return parseCompressedArgumentEntryWithFirst(parser, firstArgument, arguments);
+}
+
+inline void applyArgumentTypes(ArrayRef<Type> inputTypes, SmallVectorImpl<OpAsmParser::Argument>& arguments) {
+  for (auto [argument, inputType] : llvm::zip_equal(arguments, inputTypes))
+    argument.type = inputType;
+}
+
+inline ParseResult parseArgumentBindings(OpAsmParser& parser,
+                                         SmallVectorImpl<OpAsmParser::Argument>& arguments,
+                                         SmallVectorImpl<OpAsmParser::UnresolvedOperand>& operands) {
+  if (succeeded(parser.parseOptionalLParen())) {
+    if (succeeded(parser.parseOptionalRParen())) {
+      if (parser.parseEqual() || parseCompressedOperandList(parser, ListDelimiter::Paren, operands))
+        return failure();
+      return success();
+    }
+
+    OpAsmParser::Argument firstArgument;
+    if (parser.parseArgument(firstArgument) || parseCompressedArgumentEntryWithFirst(parser, firstArgument, arguments))
+      return failure();
+    while (succeeded(parser.parseOptionalComma()))
+      if (parseOneCompressedArgumentEntry(parser, arguments))
+        return failure();
+    if (parser.parseRParen() || parser.parseEqual()
+        || parseCompressedOperandList(parser, ListDelimiter::Paren, operands)) {
+      return failure();
+    }
+    return success();
+  }
+
+  OpAsmParser::Argument argument;
+  if (parser.parseArgument(argument) || parser.parseEqual()
+      || parseCompressedOperandList(parser, ListDelimiter::Paren, operands)) {
+    return failure();
+  }
+  arguments.push_back(argument);
+  return success();
+}
+
+} // namespace compact_asm
+} // namespace onnx_mlir
+
+#endif

src/PIM/Common/IR/CoreBlockUtils.cpp

@@ -0,0 +1,67 @@
+#include "mlir/Dialect/Arith/IR/Arith.h"
+#include "mlir/Dialect/SCF/IR/SCF.h"
+
+#include "src/Accelerators/PIM/Common/IR/CoreBlockUtils.hpp"
+#include "src/Accelerators/PIM/Dialect/Pim/PimOps.hpp"
+
+namespace onnx_mlir {
+
+bool isCoreStaticAddressOp(mlir::Operation* op) {
+  return mlir::isa<mlir::arith::ConstantOp,
+                   mlir::arith::AddIOp,
+                   mlir::arith::SubIOp,
+                   mlir::arith::MulIOp,
+                   mlir::arith::DivUIOp,
+                   mlir::arith::RemUIOp,
+                   mlir::arith::IndexCastOp,
+                   mlir::memref::AllocOp,
+                   mlir::memref::SubViewOp,
+                   mlir::memref::CastOp,
+                   mlir::memref::CollapseShapeOp,
+                   mlir::memref::ExpandShapeOp>(op);
+}
+
+mlir::LogicalResult
+walkPimCoreBlock(mlir::Block& block,
+                 const StaticValueKnowledge& knowledge,
+                 llvm::function_ref<mlir::LogicalResult(mlir::Operation&, const StaticValueKnowledge&)> callback) {
+  bool hasFailure = false;
+  for (mlir::Operation& op : block) {
+    if (mlir::isa<pim::PimHaltOp, mlir::scf::YieldOp>(op) || isCoreStaticAddressOp(&op))
+      continue;
+
+    if (auto forOp = mlir::dyn_cast<mlir::scf::ForOp>(op)) {
+      mlir::Block& loopBody = forOp.getRegion().front();
+      auto lowerBound = resolveIndexValue(forOp.getLowerBound(), knowledge);
+      auto upperBound = resolveIndexValue(forOp.getUpperBound(), knowledge);
+      auto step = resolveIndexValue(forOp.getStep(), knowledge);
+      if (failed(lowerBound) || failed(upperBound) || failed(step) || *step <= 0) {
+        forOp.emitOpError("requires statically evaluable scf.for bounds for PIM codegen");
+        hasFailure = true;
+        continue;
+      }
+
+      llvm::SmallVector<mlir::Value> iterValues(forOp.getInitArgs().begin(), forOp.getInitArgs().end());
+      for (int64_t inductionValue = *lowerBound; inductionValue < *upperBound; inductionValue += *step) {
+        StaticValueKnowledge loopKnowledge = knowledge;
+        loopKnowledge.indexValues[forOp.getInductionVar()] = inductionValue;
+        for (auto [iterArg, iterValue] : llvm::zip_equal(forOp.getRegionIterArgs(), iterValues))
+          loopKnowledge.aliases[iterArg] = iterValue;
+
+        if (failed(walkPimCoreBlock(loopBody, loopKnowledge, callback)))
+          hasFailure = true;
+
+        auto yieldOp = mlir::cast<mlir::scf::YieldOp>(loopBody.getTerminator());
+        for (auto [index, yieldedValue] : llvm::enumerate(yieldOp.getOperands()))
+          iterValues[index] = resolveLoopCarriedAlias(yieldedValue, loopKnowledge);
+      }
+      continue;
+    }
+
+    if (failed(callback(op, knowledge)))
+      hasFailure = true;
+  }
+  return mlir::success(!hasFailure);
+}
+
+} // namespace onnx_mlir

src/PIM/Common/IR/CoreBlockUtils.hpp

@@ -0,0 +1,24 @@
+#pragma once
+
+#include "mlir/IR/Block.h"
+#include "mlir/Support/LogicalResult.h"
+
+#include "llvm/ADT/STLFunctionalExtras.h"
+
+#include "src/Accelerators/PIM/Common/IR/AddressAnalysis.hpp"
+
+namespace onnx_mlir {
+
+/// Returns true for ops in a `pim.core` body that only participate in static
+/// address or index computation and therefore do not emit PIM instructions.
+bool isCoreStaticAddressOp(mlir::Operation* op);
+
+/// Walks a `pim.core` body, statically unrolling nested `scf.for` loops when
+/// their bounds are known and invoking `callback` only on instruction-emitting
+/// operations.
+mlir::LogicalResult
+walkPimCoreBlock(mlir::Block& block,
+                 const StaticValueKnowledge& knowledge,
+                 llvm::function_ref<mlir::LogicalResult(mlir::Operation&, const StaticValueKnowledge&)> callback);
+
+} // namespace onnx_mlir

src/PIM/Common/IR/EntryPointUtils.cpp

@@ -0,0 +1,45 @@
+#include "src/Accelerators/PIM/Common/IR/EntryPointUtils.hpp"
+#include "src/Dialect/ONNX/ONNXOps.hpp"
+
+namespace onnx_mlir {
+
+llvm::FailureOr<mlir::func::FuncOp> getPimEntryFunc(mlir::ModuleOp moduleOp) {
+  if (!moduleOp)
+    return mlir::failure();
+
+  llvm::SmallVector<mlir::ONNXEntryPointOp> entryPoints(moduleOp.getOps<mlir::ONNXEntryPointOp>());
+  if (entryPoints.size() > 1) {
+    moduleOp.emitError("PIM pipeline requires a single ONNX entry point, but found ") << entryPoints.size();
+    return mlir::failure();
+  }
+  if (!entryPoints.empty()) {
+    auto entryPointAttr =
+      entryPoints.front()->getAttrOfType<mlir::SymbolRefAttr>(mlir::ONNXEntryPointOp::getEntryPointFuncAttrName());
+    if (!entryPointAttr) {
+      entryPoints.front().emitOpError("is missing the entry point function attribute");
+      return mlir::failure();
+    }
+    auto entryFunc = moduleOp.lookupSymbol<mlir::func::FuncOp>(entryPointAttr.getLeafReference().getValue());
+    if (!entryFunc) {
+      entryPoints.front().emitOpError("references an unknown entry function ")
+        << entryPointAttr.getLeafReference().getValue();
+      return mlir::failure();
+    }
+    return entryFunc;
+  }
+
+  if (auto mainGraphFunc = moduleOp.lookupSymbol<mlir::func::FuncOp>("main_graph"))
+    return mainGraphFunc;
+
+  llvm::SmallVector<mlir::func::FuncOp> nonExternalFuncs;
+  for (auto funcOp : moduleOp.getOps<mlir::func::FuncOp>())
+    if (!funcOp.isExternal())
+      nonExternalFuncs.push_back(funcOp);
+  if (nonExternalFuncs.size() == 1)
+    return nonExternalFuncs.front();
+
+  moduleOp.emitError("could not resolve a unique PIM entry function");
+  return mlir::failure();
+}
+
+} // namespace onnx_mlir

src/PIM/Common/IR/EntryPointUtils.hpp

@@ -0,0 +1,13 @@
+#pragma once
+
+#include "mlir/Dialect/Func/IR/FuncOps.h"
+#include "mlir/IR/BuiltinOps.h"
+
+namespace onnx_mlir {
+
+/// Resolves the function the PIM pipeline should treat as its entry point.
+/// Prefers ONNX entry-point metadata, then `main_graph`, then the only
+/// non-external function if the module is otherwise unambiguous.
+llvm::FailureOr<mlir::func::FuncOp> getPimEntryFunc(mlir::ModuleOp moduleOp);
+
+} // namespace onnx_mlir

src/PIM/Common/IR/ShapeUtils.cpp

@@ -0,0 +1,89 @@
+#include "llvm/ADT/STLExtras.h"
+
+#include "src/Accelerators/PIM/Common/IR/ShapeUtils.hpp"
+
+namespace onnx_mlir {
+
+llvm::SmallVector<int64_t> computeRowMajorStrides(llvm::ArrayRef<int64_t> shape) {
+  llvm::SmallVector<int64_t> strides(shape.size(), 1);
+  for (int64_t dim = static_cast<int64_t>(shape.size()) - 2; dim >= 0; --dim)
+    strides[dim] = strides[dim + 1] * shape[dim + 1];
+  return strides;
+}
+
+llvm::SmallVector<int64_t>
+delinearizeIndex(int64_t linearIndex, llvm::ArrayRef<int64_t> shape, llvm::ArrayRef<int64_t> strides) {
+  llvm::SmallVector<int64_t> indices(shape.size(), 0);
+  for (auto [dim, stride] : llvm::enumerate(strides)) {
+    indices[dim] = linearIndex / stride;
+    linearIndex %= stride;
+  }
+  return indices;
+}
+
+int64_t linearizeIndex(llvm::ArrayRef<int64_t> indices, llvm::ArrayRef<int64_t> strides) {
+  int64_t linearIndex = 0;
+  for (auto [index, stride] : llvm::zip_equal(indices, strides))
+    linearIndex += index * stride;
+  return linearIndex;
+}
+
+int64_t getNumElements(llvm::ArrayRef<int64_t> shape) {
+  int64_t numElements = 1;
+  for (int64_t dim : shape)
+    numElements *= dim;
+  return numElements;
+}
+
+bool isMemoryContiguous(llvm::ArrayRef<int64_t> srcShape,
+                        llvm::ArrayRef<int64_t> offsets,
+                        llvm::ArrayRef<int64_t> sizes,
+                        llvm::ArrayRef<int64_t> strides) {
+  if (std::any_of(strides.begin(), strides.end(), [](int64_t stride) -> bool { return stride != 1; }))
+    return false;
+
+  auto offsetsAndSizesAndShape = llvm::zip_equal(llvm::make_range(offsets.rbegin(), offsets.rend()),
+                                                 llvm::make_range(sizes.rbegin(), sizes.rend()),
+                                                 llvm::make_range(srcShape.rbegin(), srcShape.rend()));
+
+  auto firstNonZeroOffset = std::find_if(
+    offsetsAndSizesAndShape.begin(), offsetsAndSizesAndShape.end(), [&](auto offsetAndSizeAndShape) -> bool {
+      auto [offset, _size, _dimension] = offsetAndSizeAndShape;
+      return offset != 0;
+    });
+
+  if (firstNonZeroOffset != offsetsAndSizesAndShape.end()) {
+    auto [offset, size, dimension] = *firstNonZeroOffset;
+    if (size > dimension - offset)
+      return false;
+    ++firstNonZeroOffset;
+
+    if (std::any_of(firstNonZeroOffset, offsetsAndSizesAndShape.end(), [](auto offsetAndSizeAndShape) -> bool {
+          auto [_offset, size, _dimension] = offsetAndSizeAndShape;
+          return size != 1;
+        }))
+      return false;
+  }
+
+  auto sizesAndShape = llvm::zip_equal(llvm::make_range(sizes.rbegin(), sizes.rend()),
+                                       llvm::make_range(srcShape.rbegin(), srcShape.rend()));
+
+  auto firstDifferentSize = std::find_if(sizesAndShape.begin(), sizesAndShape.end(), [&](auto sizeAndShape) -> bool {
+    auto [size, dimension] = sizeAndShape;
+    return size != dimension;
+  });
+
+  if (firstDifferentSize != sizesAndShape.end()) {
+    ++firstDifferentSize;
+
+    if (std::any_of(firstDifferentSize, sizesAndShape.end(), [](auto sizeAndShape) -> bool {
+          auto [size, _dimension] = sizeAndShape;
+          return size != 1;
+        }))
+      return false;
+  }
+
+  return true;
+}
+
+} // namespace onnx_mlir

src/PIM/Common/IR/ShapeUtils.hpp

@@ -0,0 +1,22 @@
+#pragma once
+
+#include "llvm/ADT/ArrayRef.h"
+#include "llvm/ADT/SmallVector.h"
+
+namespace onnx_mlir {
+
+llvm::SmallVector<int64_t> computeRowMajorStrides(llvm::ArrayRef<int64_t> shape);
+
+llvm::SmallVector<int64_t>
+delinearizeIndex(int64_t linearIndex, llvm::ArrayRef<int64_t> shape, llvm::ArrayRef<int64_t> strides);
+
+int64_t linearizeIndex(llvm::ArrayRef<int64_t> indices, llvm::ArrayRef<int64_t> strides);
+
+int64_t getNumElements(llvm::ArrayRef<int64_t> shape);
+
+bool isMemoryContiguous(llvm::ArrayRef<int64_t> srcShape,
+                        llvm::ArrayRef<int64_t> offsets,
+                        llvm::ArrayRef<int64_t> sizes,
+                        llvm::ArrayRef<int64_t> strides);
+
+} // namespace onnx_mlir

src/PIM/Common/IR/WeightUtils.cpp

@@ -0,0 +1,108 @@
+#include "mlir/Dialect/Tensor/IR/Tensor.h"
+
+#include "llvm/ADT/SmallPtrSet.h"
+#include "llvm/ADT/SmallSet.h"
+
+#include "src/Accelerators/PIM/Common/IR/WeightUtils.hpp"
+#include "src/Accelerators/PIM/Dialect/Pim/PimOps.hpp"
+#include "src/Accelerators/PIM/Dialect/Spatial/SpatialOps.hpp"
+#include "src/Dialect/ONNX/ONNXOps.hpp"
+
+namespace onnx_mlir {
+
+bool hasWeightAlways(mlir::Operation* op) { return op && op->getAttr(PimWeightAlwaysAttrName) != nullptr; }
+
+void markWeightAlways(mlir::Operation* op) {
+  assert(op && "expected valid op");
+  op->setAttr(PimWeightAlwaysAttrName, mlir::UnitAttr::get(op->getContext()));
+}
+
+namespace {
+
+template <typename MVMOpTy, typename VMMOpTy, typename ParentOpTy>
+bool hasMvmVmmWeightUse(ParentOpTy parentOp, unsigned weightIndex) {
+  bool found = false;
+  parentOp.walk([&](mlir::Operation* op) {
+    if (auto mvmOp = mlir::dyn_cast<MVMOpTy>(op))
+      found |= mvmOp.getWeightIndex() == weightIndex;
+    else if (auto vmmOp = mlir::dyn_cast<VMMOpTy>(op))
+      found |= vmmOp.getWeightIndex() == weightIndex;
+  });
+  return found;
+}
+
+template <typename MVMOpTy, typename VMMOpTy, typename ParentOpTy>
+void walkMvmVmmWeightUses(ParentOpTy parentOp, llvm::function_ref<void(mlir::OpOperand&)> callback) {
+  auto weights = parentOp.getWeights();
+  llvm::SmallSet<unsigned, 8> visited;
+  auto walkWeightIndex = [&](unsigned weightIndex) {
+    if (weightIndex < weights.size() && visited.insert(weightIndex).second)
+      callback(parentOp->getOpOperand(weightIndex));
+  };
+
+  parentOp.walk([&](MVMOpTy op) { walkWeightIndex(op.getWeightIndex()); });
+  parentOp.walk([&](VMMOpTy op) { walkWeightIndex(op.getWeightIndex()); });
+}
+
+} // namespace
+
+bool isSpatialMvmVmmWeightUse(mlir::OpOperand& use) {
+  mlir::Operation* user = use.getOwner();
+  unsigned operandIndex = use.getOperandNumber();
+
+  auto computeOp = mlir::dyn_cast<spatial::SpatCompute>(user);
+  if (!computeOp || operandIndex >= computeOp.getWeights().size())
+    return false;
+
+  return hasMvmVmmWeightUse<spatial::SpatMVMOp, spatial::SpatVMMOp>(computeOp, operandIndex);
+}
+
+bool hasOnlySpatialMvmVmmWeightUses(mlir::Value value) {
+  llvm::SmallPtrSet<mlir::Value, 8> visited;
+  auto walkUses = [&](mlir::Value currentValue, auto& self) -> bool {
+    if (!visited.insert(currentValue).second)
+      return true;
+    if (currentValue.use_empty())
+      return false;
+
+    return llvm::all_of(currentValue.getUses(), [&](mlir::OpOperand& use) {
+      if (isSpatialMvmVmmWeightUse(use))
+        return true;
+
+      mlir::Operation* user = use.getOwner();
+      if (auto extractSliceOp = mlir::dyn_cast<mlir::tensor::ExtractSliceOp>(user))
+        return extractSliceOp.getSource() == currentValue && self(extractSliceOp.getResult(), self);
+      if (auto expandShapeOp = mlir::dyn_cast<mlir::tensor::ExpandShapeOp>(user))
+        return expandShapeOp.getSrc() == currentValue && self(expandShapeOp.getResult(), self);
+      if (auto collapseShapeOp = mlir::dyn_cast<mlir::tensor::CollapseShapeOp>(user))
+        return collapseShapeOp.getSrc() == currentValue && self(collapseShapeOp.getResult(), self);
+      if (auto transposeOp = mlir::dyn_cast<mlir::ONNXTransposeOp>(user))
+        return transposeOp.getData() == currentValue && self(transposeOp.getResult(), self);
+
+      return false;
+    });
+  };
+
+  return walkUses(value, walkUses);
+}
+
+void walkPimMvmVmmWeightUses(mlir::Operation* root, llvm::function_ref<void(mlir::OpOperand&)> callback) {
+  assert(root && "expected valid root op");
+  root->walk([&](pim::PimCoreOp coreOp) {
+    coreOp.walk([&](pim::PimVMMOp vmmOp) {
+      auto weights = coreOp.getWeights();
+      unsigned weightIndex = vmmOp.getWeightIndex();
+      if (weightIndex < weights.size())
+        callback(coreOp->getOpOperand(weightIndex));
+    });
+  });
+  root->walk([&](pim::PimCoreBatchOp coreBatchOp) {
+    auto weights = coreBatchOp.getWeights();
+    for (auto weight : weights)
+      for (mlir::OpOperand& use : weight.getUses())
+        if (use.getOwner() == coreBatchOp.getOperation())
+          callback(use);
+  });
+}
+
+} // namespace onnx_mlir

src/PIM/Common/IR/WeightUtils.hpp

@@ -0,0 +1,29 @@
+#pragma once
+
+#include "mlir/IR/Operation.h"
+#include "mlir/IR/Value.h"
+
+#include "llvm/ADT/STLFunctionalExtras.h"
+#include "llvm/ADT/StringRef.h"
+
+inline constexpr llvm::StringRef PimWeightAlwaysAttrName = "weightAlways";
+
+namespace onnx_mlir {
+
+bool hasWeightAlways(mlir::Operation* op);
+
+/// Tags an op as producing a value that should stay materialized as a reusable
+/// weight across later PIM lowering/codegen stages.
+void markWeightAlways(mlir::Operation* op);
+
+bool isSpatialMvmVmmWeightUse(mlir::OpOperand& use);
+
+/// Returns true when a value flows only into Spatial weighted MVM/VMM operands,
+/// allowing later passes to preserve it as a dedicated weight-like object.
+bool hasOnlySpatialMvmVmmWeightUses(mlir::Value value);
+
+/// Visits weight operands consumed by Pim core ops/core batches so downstream
+/// passes can identify globals that must remain weight-backed.
+void walkPimMvmVmmWeightUses(mlir::Operation* root, llvm::function_ref<void(mlir::OpOperand&)> callback);
+
+} // namespace onnx_mlir

src/PIM/Common/PimCommon.cpp

@@ -1,546 +0,0 @@
-#include "mlir/Dialect/Arith/IR/Arith.h"
-#include "mlir/Dialect/MemRef/IR/MemRef.h"
-#include "mlir/Dialect/SCF/IR/SCF.h"
-#include "mlir/IR/BuiltinTypeInterfaces.h"
-#include "mlir/Interfaces/DestinationStyleOpInterface.h"
-
-#include "llvm/Support/raw_os_ostream.h"
-
-#include <filesystem>
-#include <fstream>
-
-#include "src/Accelerators/PIM/Common/PimCommon.hpp"
-#include "src/Accelerators/PIM/Dialect/Pim/PimOps.hpp"
-#include "src/Accelerators/PIM/Dialect/Spatial/SpatialOps.hpp"
-#include "src/Compiler/CompilerOptions.hpp"
-#include "src/Dialect/ONNX/ONNXOps.hpp"
-
-using namespace mlir;
-
-namespace onnx_mlir {
-
-std::string getOutputDir() {
-  if (outputBaseName.empty() || outputBaseName == "-")
-    return {};
-
-  size_t lastSlash = outputBaseName.find_last_of('/');
-  if (lastSlash == std::string::npos)
-    return ".";
-  return outputBaseName.substr(0, lastSlash);
-}
-
-void createDirectory(const std::string& directory) {
-  std::error_code errorCode;
-  std::filesystem::create_directories(directory, errorCode);
-  assert(!errorCode && ("Failed to create directory: " + errorCode.message()).data());
-}
-
-void dumpModule(ModuleOp moduleOp, const std::string& name) {
-  std::string outputDir = getOutputDir();
-  if (outputDir.empty())
-    return;
-
-  std::string dialectsDir = outputDir + "/dialects";
-  createDirectory(dialectsDir);
-
-  std::fstream file(dialectsDir + "/" + name + ".mlir", std::ios::out);
-  llvm::raw_os_ostream os(file);
-  os << *moduleOp;
-  os.flush();
-  file.close();
-}
-
-FailureOr<func::FuncOp> getPimEntryFunc(ModuleOp moduleOp) {
-  if (!moduleOp)
-    return failure();
-
-  SmallVector<ONNXEntryPointOp> entryPoints(moduleOp.getOps<ONNXEntryPointOp>());
-  if (entryPoints.size() > 1) {
-    moduleOp.emitError("PIM pipeline requires a single ONNX entry point, but found ") << entryPoints.size();
-    return failure();
-  }
-  if (!entryPoints.empty()) {
-    auto entryPointAttr =
-      entryPoints.front()->getAttrOfType<SymbolRefAttr>(ONNXEntryPointOp::getEntryPointFuncAttrName());
-    if (!entryPointAttr) {
-      entryPoints.front().emitOpError("is missing the entry point function attribute");
-      return failure();
-    }
-    auto entryFunc = moduleOp.lookupSymbol<func::FuncOp>(entryPointAttr.getLeafReference().getValue());
-    if (!entryFunc) {
-      entryPoints.front().emitOpError("references an unknown entry function ")
-        << entryPointAttr.getLeafReference().getValue();
-      return failure();
-    }
-    return entryFunc;
-  }
-
-  if (auto mainGraphFunc = moduleOp.lookupSymbol<func::FuncOp>("main_graph"))
-    return mainGraphFunc;
-
-  SmallVector<func::FuncOp> nonExternalFuncs;
-  for (auto funcOp : moduleOp.getOps<func::FuncOp>())
-    if (!funcOp.isExternal())
-      nonExternalFuncs.push_back(funcOp);
-  if (nonExternalFuncs.size() == 1)
-    return nonExternalFuncs.front();
-
-  moduleOp.emitError("could not resolve a unique PIM entry function");
-  return failure();
-}
-
-bool hasWeightAlways(Operation* op) { return op && op->getAttr(PimWeightAlwaysAttrName) != nullptr; }
-
-void markWeightAlways(Operation* op) {
-  assert(op && "expected valid op");
-  op->setAttr(PimWeightAlwaysAttrName, UnitAttr::get(op->getContext()));
-}
-
-memref::GlobalOp lookupGlobalForGetGlobal(ModuleOp moduleOp, memref::GetGlobalOp getGlobalOp) {
-  if (!moduleOp || !getGlobalOp)
-    return {};
-  return moduleOp.lookupSymbol<memref::GlobalOp>(getGlobalOp.getName());
-}
-
-FailureOr<Operation*> getOtherEndOfChannel(Operation* op, bool opIsReceive, RewriterBase& rewriter) {
-
-  auto channelNewOp = op->getOperand(0).getDefiningOp<spatial::SpatChannelNewOp>();
-  if (!channelNewOp) {
-    op->emitError("User of Channel must have the first operand created by ChannelNewOp.");
-    return failure();
-  }
-  // channelNewOp should have two users: `op` and a
-  // `ChannelSendOp`/`ChannelReceiveOp`
-  auto channelUsers = channelNewOp->getUsers();
-  auto usersIterator = channelUsers.begin();
-  auto firstUser = *usersIterator;
-  usersIterator++;
-  if (usersIterator == channelUsers.end()) {
-    op->emitError("Operand generated by ChannelNewOp must have two users, "
-                  "only one found.");
-    channelNewOp->dump();
-    op->dump();
-    channelNewOp->getParentOp()->dump();
-    return failure();
-  }
-  auto secondUser = *usersIterator;
-  usersIterator++;
-  if (usersIterator != channelUsers.end()) {
-    op->emitError("Operand generated by ChannelNewOp must have two users, "
-                  "more than two found.");
-    return failure();
-  }
-  Operation* notOpUser;
-  if (firstUser == op) {
-    notOpUser = secondUser;
-  }
-  else if (secondUser == op) {
-    notOpUser = firstUser;
-  }
-  else {
-    op->emitError("Operand generated by ChannelNewOp must have two users, "
-                  "and one of them must be me, but"
-                  "none of them is actually me.");
-    return failure();
-  }
-
-  if (opIsReceive) {
-    if (!isa<spatial::SpatChannelSendOp>(notOpUser)) {
-      op->emitError("Operand generated by ChannelNewOp has two user, one is "
-                    "me, the other is not a ChannelSendOp.");
-      return failure();
-    }
-    return notOpUser;
-  }
-  else {
-    if (!isa<spatial::SpatChannelReceiveOp>(notOpUser)) {
-      op->emitError("Operand generated by ChannelNewOp has two user, one is "
-                    "me, the other is not a ChannelReceiveOp.");
-      return failure();
-    }
-    return notOpUser;
-  }
-}
-
-SmallVector<int64_t> computeRowMajorStrides(ArrayRef<int64_t> shape) {
-  SmallVector<int64_t> strides(shape.size(), 1);
-  for (int64_t dim = static_cast<int64_t>(shape.size()) - 2; dim >= 0; --dim)
-    strides[dim] = strides[dim + 1] * shape[dim + 1];
-  return strides;
-}
-
-SmallVector<int64_t> delinearizeIndex(int64_t linearIndex, ArrayRef<int64_t> shape, ArrayRef<int64_t> strides) {
-  SmallVector<int64_t> indices(shape.size(), 0);
-  for (auto [dim, stride] : llvm::enumerate(strides)) {
-    indices[dim] = linearIndex / stride;
-    linearIndex %= stride;
-  }
-  return indices;
-}
-
-int64_t linearizeIndex(ArrayRef<int64_t> indices, ArrayRef<int64_t> strides) {
-  int64_t linearIndex = 0;
-  for (auto [index, stride] : llvm::zip_equal(indices, strides))
-    linearIndex += index * stride;
-  return linearIndex;
-}
-
-int64_t getNumElements(ArrayRef<int64_t> shape) {
-  int64_t numElements = 1;
-  for (int64_t dim : shape)
-    numElements *= dim;
-  return numElements;
-}
-
-bool isMemoryContiguous(ArrayRef<int64_t> srcShape,
-                        ArrayRef<int64_t> offsets,
-                        ArrayRef<int64_t> sizes,
-                        ArrayRef<int64_t> strides) {
-  if (std::any_of(strides.begin(), strides.end(), [](int64_t stride) -> bool { return stride != 1; }))
-    return false;
-
-  auto offsetsAndSizesAndShape = llvm::zip_equal(llvm::make_range(offsets.rbegin(), offsets.rend()),
-                                                 llvm::make_range(sizes.rbegin(), sizes.rend()),
-                                                 llvm::make_range(srcShape.rbegin(), srcShape.rend()));
-
-  auto firstNonZeroOffset = std::find_if(
-    offsetsAndSizesAndShape.begin(), offsetsAndSizesAndShape.end(), [&](auto offsetAndSizeAndShape) -> bool {
-      auto [offset, _size, _dimension] = offsetAndSizeAndShape;
-      return offset != 0;
-    });
-
-  if (firstNonZeroOffset != offsetsAndSizesAndShape.end()) {
-    auto [offset, size, dimension] = *firstNonZeroOffset;
-    if (size > dimension - offset)
-      return false;
-    ++firstNonZeroOffset;
-
-    if (std::any_of(firstNonZeroOffset, offsetsAndSizesAndShape.end(), [](auto offsetAndSizeAndShape) -> bool {
-          auto [_offset, size, _dimension] = offsetAndSizeAndShape;
-          return size != 1;
-        }))
-      return false;
-  }
-
-  auto sizesAndShape = llvm::zip_equal(llvm::make_range(sizes.rbegin(), sizes.rend()),
-                                       llvm::make_range(srcShape.rbegin(), srcShape.rend()));
-
-  auto firstDifferentSize = std::find_if(sizesAndShape.begin(), sizesAndShape.end(), [&](auto sizeAndShape) -> bool {
-    auto [size, dimension] = sizeAndShape;
-    return size != dimension;
-  });
-
-  if (firstDifferentSize != sizesAndShape.end()) {
-    ++firstDifferentSize;
-
-    if (std::any_of(firstDifferentSize, sizesAndShape.end(), [](auto sizeAndShape) -> bool {
-          auto [size, _dimension] = sizeAndShape;
-          return size != 1;
-        }))
-      return false;
-  }
-
-  return true;
-}
-
-static Value resolveAlias(Value value, const StaticValueKnowledge* knowledge) {
-  if (!knowledge)
-    return value;
-
-  auto iter = knowledge->aliases.find(value);
-  while (iter != knowledge->aliases.end()) {
-    value = iter->second;
-    iter = knowledge->aliases.find(value);
-  }
-  return value;
-}
-
-// Walks through view-like ops and DPS tied operands to find the "underlying" memref value
-// behind an scf.for iter-arg. Used both when resolving a contiguous address inside a loop
-// and when propagating yielded values across iterations during static unrolling.
-static Value resolveLoopCarriedAliasImpl(Value value, const StaticValueKnowledge* knowledge) {
-  value = resolveAlias(value, knowledge);
-
-  if (auto blockArgument = dyn_cast<BlockArgument>(value))
-    return value;
-
-  Operation* definingOp = value.getDefiningOp();
-  if (!definingOp)
-    return value;
-
-  if (auto dpsDefiningOp = dyn_cast<DestinationStyleOpInterface>(definingOp)) {
-    if (auto result = dyn_cast<OpResult>(value))
-      if (OpOperand* tiedOperand = dpsDefiningOp.getTiedOpOperand(result))
-        return resolveLoopCarriedAliasImpl(tiedOperand->get(), knowledge);
-  }
-
-  if (auto castOp = dyn_cast<memref::CastOp>(definingOp))
-    return resolveLoopCarriedAliasImpl(castOp.getSource(), knowledge);
-  if (auto collapseOp = dyn_cast<memref::CollapseShapeOp>(definingOp))
-    return resolveLoopCarriedAliasImpl(collapseOp.getSrc(), knowledge);
-  if (auto expandOp = dyn_cast<memref::ExpandShapeOp>(definingOp))
-    return resolveLoopCarriedAliasImpl(expandOp.getSrc(), knowledge);
-
-  return value;
-}
-
-static FailureOr<int64_t> resolveOpFoldResult(OpFoldResult ofr, const StaticValueKnowledge* knowledge);
-
-static FailureOr<int64_t> resolveIndexValueImpl(Value value, const StaticValueKnowledge* knowledge) {
-  value = resolveAlias(value, knowledge);
-
-  if (knowledge) {
-    auto iter = knowledge->indexValues.find(value);
-    if (iter != knowledge->indexValues.end())
-      return iter->second;
-  }
-
-  auto constantOp = value.getDefiningOp<arith::ConstantOp>();
-  if (constantOp) {
-    if (auto integerAttr = dyn_cast<IntegerAttr>(constantOp.getValue()))
-      return integerAttr.getInt();
-  }
-
-  Operation* definingOp = value.getDefiningOp();
-  if (!definingOp)
-    return failure();
-
-  if (auto indexCastOp = dyn_cast<arith::IndexCastOp>(definingOp))
-    return resolveIndexValueImpl(indexCastOp.getIn(), knowledge);
-
-  if (auto addOp = dyn_cast<arith::AddIOp>(definingOp)) {
-    auto lhs = resolveIndexValueImpl(addOp.getLhs(), knowledge);
-    auto rhs = resolveIndexValueImpl(addOp.getRhs(), knowledge);
-    if (failed(lhs) || failed(rhs))
-      return failure();
-    return *lhs + *rhs;
-  }
-
-  if (auto subOp = dyn_cast<arith::SubIOp>(definingOp)) {
-    auto lhs = resolveIndexValueImpl(subOp.getLhs(), knowledge);
-    auto rhs = resolveIndexValueImpl(subOp.getRhs(), knowledge);
-    if (failed(lhs) || failed(rhs))
-      return failure();
-    return *lhs - *rhs;
-  }
-
-  if (auto mulOp = dyn_cast<arith::MulIOp>(definingOp)) {
-    auto lhs = resolveIndexValueImpl(mulOp.getLhs(), knowledge);
-    auto rhs = resolveIndexValueImpl(mulOp.getRhs(), knowledge);
-    if (failed(lhs) || failed(rhs))
-      return failure();
-    return *lhs * *rhs;
-  }
-
-  if (auto divOp = dyn_cast<arith::DivUIOp>(definingOp)) {
-    auto lhs = resolveIndexValueImpl(divOp.getLhs(), knowledge);
-    auto rhs = resolveIndexValueImpl(divOp.getRhs(), knowledge);
-    if (failed(lhs) || failed(rhs) || *rhs == 0)
-      return failure();
-    return static_cast<int64_t>(static_cast<uint64_t>(*lhs) / static_cast<uint64_t>(*rhs));
-  }
-
-  if (auto remOp = dyn_cast<arith::RemUIOp>(definingOp)) {
-    auto lhs = resolveIndexValueImpl(remOp.getLhs(), knowledge);
-    auto rhs = resolveIndexValueImpl(remOp.getRhs(), knowledge);
-    if (failed(lhs) || failed(rhs) || *rhs == 0)
-      return failure();
-    return static_cast<int64_t>(static_cast<uint64_t>(*lhs) % static_cast<uint64_t>(*rhs));
-  }
-
-  return failure();
-}
-
-static FailureOr<int64_t> resolveOpFoldResult(OpFoldResult ofr, const StaticValueKnowledge* knowledge) {
-  if (auto attr = dyn_cast<Attribute>(ofr)) {
-    auto integerAttr = dyn_cast<IntegerAttr>(attr);
-    if (!integerAttr)
-      return failure();
-    return integerAttr.getInt();
-  }
-
-  return resolveIndexValueImpl(cast<Value>(ofr), knowledge);
-}
-
-static FailureOr<ResolvedContiguousAddress> resolveContiguousAddressImpl(Value value,
-                                                                         const StaticValueKnowledge* knowledge) {
-  int64_t byteOffset = 0;
-  value = resolveAlias(value, knowledge);
-
-  while (true) {
-    if (isa<BlockArgument>(value))
-      return ResolvedContiguousAddress {value, byteOffset};
-
-    Operation* definingOp = value.getDefiningOp();
-    if (!definingOp)
-      return failure();
-
-    if (auto dpsDefiningOp = dyn_cast<DestinationStyleOpInterface>(definingOp)) {
-      OpOperand* tiedOperand = dpsDefiningOp.getTiedOpOperand(dyn_cast<OpResult>(value));
-      if (!tiedOperand)
-        return failure();
-      value = resolveAlias(tiedOperand->get(), knowledge);
-      continue;
-    }
-
-    if (auto forOp = dyn_cast<scf::ForOp>(definingOp)) {
-      auto result = dyn_cast<OpResult>(value);
-      if (!result)
-        return failure();
-
-      // Trace the loop carry back to its underlying memref, then if that memref is the
-      // loop's own iter-arg we know the base comes from the corresponding init arg
-      // (every iteration yields the same backing memory in the DPS sense).
-      auto yieldOp = cast<scf::YieldOp>(forOp.getBody()->getTerminator());
-      Value yieldedValue = resolveLoopCarriedAliasImpl(yieldOp.getOperand(result.getResultNumber()), knowledge);
-      if (auto blockArgument = dyn_cast<BlockArgument>(yieldedValue)) {
-        if (blockArgument.getOwner() == forOp.getBody() && blockArgument.getArgNumber() > 0
-            && static_cast<unsigned>(blockArgument.getArgNumber() - 1) < forOp.getInitArgs().size()) {
-          value = resolveAlias(forOp.getInitArgs()[blockArgument.getArgNumber() - 1], knowledge);
-          continue;
-        }
-      }
-
-      value = yieldedValue;
-      continue;
-    }
-
-    if (auto subviewOp = dyn_cast<memref::SubViewOp>(definingOp)) {
-      auto sourceType = dyn_cast<MemRefType>(subviewOp.getSource().getType());
-      auto subviewType = dyn_cast<MemRefType>(subviewOp.getType());
-      if (!sourceType || !subviewType || !sourceType.hasStaticShape() || !subviewType.hasStaticShape())
-        return failure();
-
-      SmallVector<int64_t> offsets;
-      SmallVector<int64_t> sizes;
-      SmallVector<int64_t> strides;
-      offsets.reserve(subviewOp.getMixedOffsets().size());
-      sizes.reserve(subviewOp.getMixedSizes().size());
-      strides.reserve(subviewOp.getMixedStrides().size());
-
-      for (OpFoldResult offset : subviewOp.getMixedOffsets()) {
-        auto resolvedOffset = resolveOpFoldResult(offset, knowledge);
-        if (failed(resolvedOffset))
-          return failure();
-        offsets.push_back(*resolvedOffset);
-      }
-
-      for (OpFoldResult size : subviewOp.getMixedSizes()) {
-        auto resolvedSize = resolveOpFoldResult(size, knowledge);
-        if (failed(resolvedSize))
-          return failure();
-        sizes.push_back(*resolvedSize);
-      }
-
-      for (OpFoldResult stride : subviewOp.getMixedStrides()) {
-        auto resolvedStride = resolveOpFoldResult(stride, knowledge);
-        if (failed(resolvedStride))
-          return failure();
-        strides.push_back(*resolvedStride);
-      }
-
-      if (!isMemoryContiguous(sourceType.getShape(), offsets, sizes, strides))
-        return failure();
-
-      auto sourceStrides = computeRowMajorStrides(sourceType.getShape());
-      byteOffset += linearizeIndex(offsets, sourceStrides) * subviewType.getElementTypeBitWidth() / 8;
-      value = resolveAlias(subviewOp.getSource(), knowledge);
-      continue;
-    }
-
-    if (auto castOp = dyn_cast<memref::CastOp>(definingOp)) {
-      value = resolveAlias(castOp.getSource(), knowledge);
-      continue;
-    }
-    if (auto collapseOp = dyn_cast<memref::CollapseShapeOp>(definingOp)) {
-      value = resolveAlias(collapseOp.getSrc(), knowledge);
-      continue;
-    }
-    if (auto expandOp = dyn_cast<memref::ExpandShapeOp>(definingOp)) {
-      value = resolveAlias(expandOp.getSrc(), knowledge);
-      continue;
-    }
-
-    if (isa<memref::AllocOp, memref::GetGlobalOp>(definingOp))
-      return ResolvedContiguousAddress {value, byteOffset};
-
-    return failure();
-  }
-}
-
-FailureOr<int64_t> resolveIndexValue(Value value) { return resolveIndexValueImpl(value, nullptr); }
-
-FailureOr<int64_t> resolveIndexValue(Value value, const StaticValueKnowledge& knowledge) {
-  return resolveIndexValueImpl(value, &knowledge);
-}
-
-FailureOr<ResolvedContiguousAddress> resolveContiguousAddress(Value value) {
-  return resolveContiguousAddressImpl(value, nullptr);
-}
-
-FailureOr<ResolvedContiguousAddress> resolveContiguousAddress(Value value, const StaticValueKnowledge& knowledge) {
-  return resolveContiguousAddressImpl(value, &knowledge);
-}
-
-Value resolveLoopCarriedAlias(Value value, const StaticValueKnowledge& knowledge) {
-  return resolveLoopCarriedAliasImpl(value, &knowledge);
-}
-
-bool isCoreStaticAddressOp(Operation* op) {
-  return isa<arith::ConstantOp,
-             arith::AddIOp,
-             arith::SubIOp,
-             arith::MulIOp,
-             arith::DivUIOp,
-             arith::RemUIOp,
-             arith::IndexCastOp,
-             memref::AllocOp,
-             memref::SubViewOp,
-             memref::CastOp,
-             memref::CollapseShapeOp,
-             memref::ExpandShapeOp>(op);
-}
-
-LogicalResult walkPimCoreBlock(Block& block,
-                               const StaticValueKnowledge& knowledge,
-                               llvm::function_ref<LogicalResult(Operation&, const StaticValueKnowledge&)> callback) {
-  bool hasFailure = false;
-  for (Operation& op : block) {
-    if (isa<pim::PimHaltOp, scf::YieldOp>(op) || isCoreStaticAddressOp(&op))
-      continue;
-
-    if (auto forOp = dyn_cast<scf::ForOp>(op)) {
-      Block& loopBody = forOp.getRegion().front();
-      auto lowerBound = resolveIndexValue(forOp.getLowerBound(), knowledge);
-      auto upperBound = resolveIndexValue(forOp.getUpperBound(), knowledge);
-      auto step = resolveIndexValue(forOp.getStep(), knowledge);
-      if (failed(lowerBound) || failed(upperBound) || failed(step) || *step <= 0) {
-        forOp.emitOpError("requires statically evaluable scf.for bounds for PIM codegen");
-        hasFailure = true;
-        continue;
-      }
-
-      SmallVector<Value> iterValues(forOp.getInitArgs().begin(), forOp.getInitArgs().end());
-      for (int64_t inductionValue = *lowerBound; inductionValue < *upperBound; inductionValue += *step) {
-        StaticValueKnowledge loopKnowledge = knowledge;
-        loopKnowledge.indexValues[forOp.getInductionVar()] = inductionValue;
-        for (auto [iterArg, iterValue] : llvm::zip_equal(forOp.getRegionIterArgs(), iterValues))
-          loopKnowledge.aliases[iterArg] = iterValue;
-
-        if (failed(walkPimCoreBlock(loopBody, loopKnowledge, callback)))
-          hasFailure = true;
-
-        auto yieldOp = cast<scf::YieldOp>(loopBody.getTerminator());
-        for (auto [index, yieldedValue] : llvm::enumerate(yieldOp.getOperands()))
-          iterValues[index] = resolveLoopCarriedAlias(yieldedValue, loopKnowledge);
-      }
-      continue;
-    }
-
-    if (failed(callback(op, knowledge)))
-      hasFailure = true;
-  }
-  return success(!hasFailure);
-}
-
-} // namespace onnx_mlir

src/PIM/Common/PimCommon.hpp

@@ -7,82 +7,22 @@
 #include "mlir/IR/Value.h"
 
 #include "llvm/ADT/DenseMap.h"
+#include "llvm/ADT/STLFunctionalExtras.h"
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/ADT/StringRef.h"
 
+#include "src/Accelerators/PIM/Common/IR/AddressAnalysis.hpp"
+#include "src/Accelerators/PIM/Common/IR/CoreBlockUtils.hpp"
+#include "src/Accelerators/PIM/Common/IR/EntryPointUtils.hpp"
+#include "src/Accelerators/PIM/Common/IR/ShapeUtils.hpp"
+#include "src/Accelerators/PIM/Common/IR/WeightUtils.hpp"
+#include "src/Accelerators/PIM/Common/Support/DebugDump.hpp"
+#include "src/Accelerators/PIM/Common/Support/FileSystemUtils.hpp"
 #include "src/Compiler/CompilerOptions.hpp"
 
-inline constexpr llvm::StringRef PimWeightAlwaysAttrName = "weightAlways";
-
 namespace onnx_mlir {
 
-struct ResolvedContiguousAddress {
+inline constexpr llvm::StringLiteral kCoreIdAttrName = "coreId";
-  mlir::Value base;
+inline constexpr llvm::StringLiteral kCoreIdsAttrName = "coreIds";
-  int64_t byteOffset = 0;
-};
-
-struct StaticValueKnowledge {
-  llvm::DenseMap<mlir::Value, int64_t> indexValues;
-  llvm::DenseMap<mlir::Value, mlir::Value> aliases;
-
-  StaticValueKnowledge() {}
-};
-
-std::string getOutputDir();
-
-void createDirectory(const std::string& directory);
-
-void dumpModule(mlir::ModuleOp moduleOp, const std::string& name);
-
-llvm::FailureOr<mlir::func::FuncOp> getPimEntryFunc(mlir::ModuleOp moduleOp);
-
-bool hasWeightAlways(mlir::Operation* op);
-
-void markWeightAlways(mlir::Operation* op);
-
-mlir::memref::GlobalOp lookupGlobalForGetGlobal(mlir::ModuleOp moduleOp, mlir::memref::GetGlobalOp getGlobalOp);
-
-llvm::FailureOr<mlir::Operation*>
-getOtherEndOfChannel(mlir::Operation* op, bool opIsReceive, mlir::RewriterBase& rewriter);
-
-llvm::SmallVector<int64_t> computeRowMajorStrides(llvm::ArrayRef<int64_t> shape);
-
-llvm::SmallVector<int64_t>
-delinearizeIndex(int64_t linearIndex, llvm::ArrayRef<int64_t> shape, llvm::ArrayRef<int64_t> strides);
-
-int64_t linearizeIndex(llvm::ArrayRef<int64_t> indices, llvm::ArrayRef<int64_t> strides);
-
-int64_t getNumElements(llvm::ArrayRef<int64_t> shape);
-
-bool isMemoryContiguous(llvm::ArrayRef<int64_t> srcShape,
-                        llvm::ArrayRef<int64_t> offsets,
-                        llvm::ArrayRef<int64_t> sizes,
-                        llvm::ArrayRef<int64_t> strides);
-
-llvm::FailureOr<ResolvedContiguousAddress> resolveContiguousAddress(mlir::Value value);
-llvm::FailureOr<ResolvedContiguousAddress> resolveContiguousAddress(mlir::Value value,
-                                                                    const StaticValueKnowledge& knowledge);
-
-llvm::FailureOr<int64_t> resolveIndexValue(mlir::Value value);
-llvm::FailureOr<int64_t> resolveIndexValue(mlir::Value value, const StaticValueKnowledge& knowledge);
-
-/// Follows alias and view/DPS chains using `knowledge` to find the value an scf.for
-/// iter-arg is ultimately backed by. Used when interpreting scf.for loop carries.
-mlir::Value resolveLoopCarriedAlias(mlir::Value value, const StaticValueKnowledge& knowledge);
-
-/// Returns true for ops inside a pim.core body that do not emit any PIM instruction and
-/// only contribute to static addressing or index computations (arith integer math,
-/// memref view ops, memref.alloc, arith.constant).
-bool isCoreStaticAddressOp(mlir::Operation* op);
-
-/// Walks `block` (the body of a pim.core region or an scf.for nested in it), statically
-/// unrolling any scf.for with resolvable bounds using `knowledge`. For each remaining op
-/// that is not skipped (pim.halt, scf.yield, or isCoreStaticAddressOp), `callback` is
-/// invoked with the op and the in-scope knowledge. The walker keeps going after a callback
-/// failure so callers can collect multiple diagnostics, but propagates the overall result.
-mlir::LogicalResult
-walkPimCoreBlock(mlir::Block& block,
-                 const StaticValueKnowledge& knowledge,
-                 llvm::function_ref<mlir::LogicalResult(mlir::Operation&, const StaticValueKnowledge&)> callback);
 
 } // namespace onnx_mlir

src/PIM/Common/Support/DebugDump.cpp

@@ -0,0 +1,27 @@
+#include "llvm/Support/raw_os_ostream.h"
+
+#include <fstream>
+
+#include "src/Accelerators/PIM/Common/Support/DebugDump.hpp"
+#include "src/Accelerators/PIM/Common/Support/FileSystemUtils.hpp"
+
+namespace onnx_mlir {
+
+void dumpModule(mlir::ModuleOp moduleOp, const std::string& name) {
+  std::string outputDir = getOutputDir();
+  if (outputDir.empty())
+    return;
+
+  std::string dialectsDir = outputDir + "/dialects";
+  createDirectory(dialectsDir);
+
+  std::fstream file(dialectsDir + "/" + name + ".mlir", std::ios::out);
+  llvm::raw_os_ostream os(file);
+  mlir::OpPrintingFlags flags;
+  flags.elideLargeElementsAttrs();
+  moduleOp.print(os, flags);
+  os.flush();
+  file.close();
+}
+
+} // namespace onnx_mlir

src/PIM/Common/Support/DebugDump.hpp

@@ -0,0 +1,13 @@
+#pragma once
+
+#include "mlir/IR/BuiltinOps.h"
+
+#include <string>
+
+namespace onnx_mlir {
+
+/// Emits a MLIR snapshot under the current compiler output
+/// directory for pass-level debugging.
+void dumpModule(mlir::ModuleOp moduleOp, const std::string& name);
+
+} // namespace onnx_mlir

src/PIM/Common/Support/Diagnostics.cpp

@@ -0,0 +1,41 @@
+#include "llvm/ADT/STLExtras.h"
+
+#include "src/Accelerators/PIM/Common/Support/Diagnostics.hpp"
+
+namespace onnx_mlir::pim {
+
+mlir::InFlightDiagnostic emitUnsupportedStaticShapeDiagnostic(mlir::Operation* op, llvm::StringRef valueDescription) {
+  return op->emitOpError() << "requires statically shaped " << valueDescription;
+}
+
+mlir::InFlightDiagnostic emitUnsupportedRankDiagnostic(mlir::Operation* op,
+                                                       llvm::StringRef valueDescription,
+                                                       int64_t actualRank,
+                                                       llvm::ArrayRef<int64_t> supportedRanks) {
+  auto diag = op->emitOpError() << "has unsupported rank " << actualRank << " for " << valueDescription;
+  if (supportedRanks.empty())
+    return diag;
+
+  diag << "; supported rank";
+  if (supportedRanks.size() != 1)
+    diag << 's';
+  diag << ' ';
+
+  llvm::interleaveComma(supportedRanks, diag, [&](int64_t rank) { diag << rank; });
+  return diag;
+}
+
+mlir::InFlightDiagnostic
+emitMissingSymbolDiagnostic(mlir::Operation* op, llvm::StringRef symbolKind, llvm::StringRef symbolName) {
+  return op->emitOpError() << "references missing " << symbolKind << " `" << symbolName << "`";
+}
+
+mlir::LogicalResult emitFileSystemError(mlir::Location loc,
+                                        llvm::StringRef action,
+                                        llvm::StringRef path,
+                                        const std::error_code& errorCode) {
+  mlir::emitError(loc) << "failed to " << action << " `" << path << "`: " << errorCode.message();
+  return mlir::failure();
+}
+
+} // namespace onnx_mlir::pim

src/PIM/Common/Support/Diagnostics.hpp

@@ -0,0 +1,38 @@
+#pragma once
+
+#include "mlir/IR/Diagnostics.h"
+#include "mlir/IR/Operation.h"
+#include "mlir/Support/LogicalResult.h"
+
+#include "llvm/ADT/ArrayRef.h"
+#include "llvm/ADT/StringRef.h"
+
+#include <system_error>
+
+namespace onnx_mlir::pim {
+
+/// Emits a consistent diagnostic for target paths that require static shapes.
+mlir::InFlightDiagnostic emitUnsupportedStaticShapeDiagnostic(mlir::Operation* op, llvm::StringRef valueDescription);
+
+/// Emits a consistent diagnostic for unsupported ranks while listing the ranks
+/// accepted by the current lowering/codegen path.
+mlir::InFlightDiagnostic emitUnsupportedRankDiagnostic(mlir::Operation* op,
+                                                       llvm::StringRef valueDescription,
+                                                       int64_t actualRank,
+                                                       llvm::ArrayRef<int64_t> supportedRanks);
+
+/// Emits a consistent diagnostic for missing symbol/global references.
+mlir::InFlightDiagnostic
+emitMissingSymbolDiagnostic(mlir::Operation* op, llvm::StringRef symbolKind, llvm::StringRef symbolName);
+
+/// Converts a filesystem error into an MLIR failure diagnostic anchored at
+/// the relevant IR location.
+mlir::LogicalResult
+emitFileSystemError(mlir::Location loc, llvm::StringRef action, llvm::StringRef path, const std::error_code& errorCode);
+
+template <typename T>
+mlir::LogicalResult failureOrToLogicalResult(const llvm::FailureOr<T>& value) {
+  return mlir::success(succeeded(value));
+}
+
+} // namespace onnx_mlir::pim

src/PIM/Common/Support/FileSystemUtils.cpp

@@ -0,0 +1,24 @@
+#include <filesystem>
+
+#include "src/Accelerators/PIM/Common/Support/FileSystemUtils.hpp"
+#include "src/Compiler/CompilerOptions.hpp"
+
+namespace onnx_mlir {
+
+std::string getOutputDir() {
+  if (outputBaseName.empty() || outputBaseName == "-")
+    return {};
+
+  size_t lastSlash = outputBaseName.find_last_of('/');
+  if (lastSlash == std::string::npos)
+    return ".";
+  return outputBaseName.substr(0, lastSlash);
+}
+
+void createDirectory(const std::string& directory) {
+  std::error_code errorCode;
+  std::filesystem::create_directories(directory, errorCode);
+  assert(!errorCode && ("Failed to create directory: " + errorCode.message()).data());
+}
+
+} // namespace onnx_mlir

src/PIM/Common/Support/FileSystemUtils.hpp

@@ -0,0 +1,13 @@
+#pragma once
+
+#include <string>
+
+namespace onnx_mlir {
+
+/// Returns the directory that should hold PIM artifacts/debug dumps for the
+/// current compiler invocation.
+std::string getOutputDir();
+
+void createDirectory(const std::string& directory);
+
+} // namespace onnx_mlir

src/PIM/Compiler/CMakeLists.txt

@@ -15,7 +15,10 @@ add_pim_library(OMPimCompilerOptions
 
 add_pim_library(OMPimCompilerUtils
   PimCompilerUtils.cpp
+  PimArtifactWriter.cpp
+  PimBatchEmission.cpp
   PimCodeGen.cpp
+  PimWeightEmitter.cpp
 
   EXCLUDE_FROM_OM_LIBS
 

src/PIM/Compiler/PimArtifactWriter.cpp

@@ -0,0 +1,123 @@
+#include "mlir/Dialect/MemRef/IR/MemRef.h"
+
+#include "llvm/ADT/SmallPtrSet.h"
+#include "llvm/Support/FileSystem.h"
+#include "llvm/Support/raw_ostream.h"
+
+#include <algorithm>
+#include <cassert>
+#include <cstring>
+#include <vector>
+
+#include "src/Accelerators/PIM/Common/IR/WeightUtils.hpp"
+#include "src/Accelerators/PIM/Compiler/PimArtifactWriter.hpp"
+#include "src/Accelerators/PIM/Compiler/PimCodeGen.hpp"
+#include "src/Accelerators/PIM/Compiler/PimCompilerOptions.hpp"
+
+using namespace llvm;
+using namespace mlir;
+
+namespace onnx_mlir {
+
+OnnxMlirCompilerErrorCodes writeHostCoreJson(StringRef outputDirPath) {
+  std::error_code errorCode;
+  std::string outputHostCorePath = outputDirPath.str() + "/core_0.json";
+  raw_fd_ostream hostFileStream(outputHostCorePath, errorCode);
+  if (errorCode) {
+    errs() << "Error while opening host core file `" << outputHostCorePath << "`: " << errorCode.message() << '\n';
+    return InvalidOutputFileAccess;
+  }
+
+  // The host core json contains two no-op-like instructions to satisfy pimsim-nn.
+  hostFileStream << "[{\"imm\":0,\"op\":\"sldi\",\"rd\":0},{\"imm\":0,\"op\":\"sldi\",\"rd\":0}]";
+  hostFileStream.close();
+  return CompilerSuccess;
+}
+
+OnnxMlirCompilerErrorCodes
+writeMemoryBinary(ModuleOp moduleOp, func::FuncOp funcOp, PimAcceleratorMemory& memory, StringRef outputDirPath) {
+  auto memoryFilePath = (outputDirPath + "/memory.bin").str();
+  std::error_code errorCode;
+  raw_fd_ostream memoryFileStream(memoryFilePath, errorCode, sys::fs::OF_None);
+  if (errorCode) {
+    errs() << "Error while opening memory file " << memoryFilePath << ": " << errorCode.message() << '\n';
+    return InvalidOutputFileAccess;
+  }
+
+  std::vector<char> memoryBuffer(memory.hostMem.getFirstAvailableAddress(), 0);
+
+  SmallPtrSet<Operation*, 16> writtenGlobals;
+  funcOp.walk([&](memref::GetGlobalOp getGlobalOp) {
+    if (hasWeightAlways(getGlobalOp))
+      return;
+    auto globalOp = lookupGlobalForGetGlobal(moduleOp, getGlobalOp);
+    if (!globalOp)
+      return;
+    if (!writtenGlobals.insert(globalOp.getOperation()).second)
+      return;
+    auto initialValue = globalOp.getInitialValue();
+    if (!initialValue)
+      return;
+    auto denseAttr = dyn_cast<DenseElementsAttr>(*initialValue);
+    if (!denseAttr)
+      return;
+
+    MemEntry memEntry = memory.hostMem.getMemEntry(getGlobalOp.getResult());
+    ArrayRef<char> rawData = denseAttr.getRawData();
+    char* dst = memoryBuffer.data() + memEntry.address;
+
+    if (denseAttr.isSplat()) {
+      size_t elementSize = rawData.size();
+      assert(elementSize * getGlobalOp.getType().getNumElements() == memEntry.size && "Data size mismatch");
+      for (size_t offset = 0; offset < memEntry.size; offset += elementSize)
+        std::memcpy(dst + offset, rawData.data(), std::min(elementSize, memEntry.size - offset));
+    }
+    else {
+      assert(rawData.size() == memEntry.size && "Data size mismatch");
+      std::memcpy(dst, rawData.data(), rawData.size());
+    }
+  });
+
+  memoryFileStream.write(memoryBuffer.data(), memoryBuffer.size());
+  memoryFileStream.close();
+  return CompilerSuccess;
+}
+
+OnnxMlirCompilerErrorCodes writeConfigJson(func::FuncOp funcOp,
+                                           PimAcceleratorMemory& memory,
+                                           size_t maxCoreId,
+                                           json::Object xbarsPerArrayGroup,
+                                           StringRef outputDirPath) {
+  json::Object configJson;
+
+  configJson["core_cnt"] = maxCoreId + 1;
+  configJson["adc_count"] = 16;
+  configJson["cell_precision"] = 2;
+  configJson["xbar_array_count"] = crossbarCountInCore.getValue();
+  configJson["xbar_size"] = {crossbarSize.getValue(), crossbarSize.getValue()};
+  configJson["array_group_map"] = std::move(xbarsPerArrayGroup);
+
+  json::Array inputsAddresses;
+  for (BlockArgument input : funcOp.getArguments())
+    inputsAddresses.push_back(memory.getValueAddress(input));
+  configJson["inputs_addresses"] = std::move(inputsAddresses);
+
+  json::Array outputsAddresses;
+  for (func::ReturnOp returnOp : funcOp.getOps<func::ReturnOp>())
+    for (mlir::Value output : returnOp.getOperands())
+      outputsAddresses.push_back(memory.getValueAddress(output));
+  configJson["outputs_addresses"] = std::move(outputsAddresses);
+
+  auto configPath = (outputDirPath + "/config.json").str();
+  std::error_code errorCode;
+  raw_fd_ostream jsonOS(configPath, errorCode);
+  if (errorCode) {
+    errs() << "Error while opening config file: " << errorCode.message() << '\n';
+    return InvalidOutputFileAccess;
+  }
+  jsonOS << json::Value(std::move(configJson)) << '\n';
+  jsonOS.close();
+  return CompilerSuccess;
+}
+
+} // namespace onnx_mlir

src/PIM/Compiler/PimArtifactWriter.hpp

@@ -0,0 +1,26 @@
+#pragma once
+
+#include "mlir/Dialect/Func/IR/FuncOps.h"
+#include "mlir/IR/BuiltinOps.h"
+
+#include "llvm/ADT/StringRef.h"
+#include "llvm/Support/JSON.h"
+
+#include "onnx-mlir/Compiler/OMCompilerTypes.h"
+
+namespace onnx_mlir {
+
+class PimAcceleratorMemory;
+
+OnnxMlirCompilerErrorCodes writeHostCoreJson(llvm::StringRef outputDirPath);
+OnnxMlirCompilerErrorCodes writeMemoryBinary(mlir::ModuleOp moduleOp,
+                                             mlir::func::FuncOp funcOp,
+                                             PimAcceleratorMemory& memory,
+                                             llvm::StringRef outputDirPath);
+OnnxMlirCompilerErrorCodes writeConfigJson(mlir::func::FuncOp funcOp,
+                                           PimAcceleratorMemory& memory,
+                                           size_t maxCoreId,
+                                           llvm::json::Object xbarsPerArrayGroup,
+                                           llvm::StringRef outputDirPath);
+
+} // namespace onnx_mlir

src/PIM/Compiler/PimBatchEmission.cpp

@@ -0,0 +1,126 @@
+#include "mlir/IR/Builders.h"
+#include "mlir/IR/BuiltinOps.h"
+#include "mlir/IR/IRMapping.h"
+
+#include "src/Accelerators/PIM/Common/PimCommon.hpp"
+#include "src/Accelerators/PIM/Compiler/PimBatchEmission.hpp"
+
+using namespace mlir;
+
+namespace onnx_mlir {
+namespace {
+
+static SmallVector<int32_t> getBatchCoreIds(pim::PimCoreBatchOp coreBatchOp) {
+  auto coreIdsAttr = coreBatchOp->getAttrOfType<DenseI32ArrayAttr>(onnx_mlir::kCoreIdsAttrName);
+  assert(coreIdsAttr && "pim.core_batch requires coreIds array attribute");
+  return SmallVector<int32_t>(coreIdsAttr.asArrayRef().begin(), coreIdsAttr.asArrayRef().end());
+}
+
+static SmallVector<int32_t> getLaneChunkCoreIds(ArrayRef<int32_t> coreIds, size_t laneCount, unsigned lane) {
+  SmallVector<int32_t> laneCoreIds;
+  laneCoreIds.reserve(coreIds.size() / laneCount);
+  for (size_t chunkIndex = 0; chunkIndex < coreIds.size() / laneCount; ++chunkIndex)
+    laneCoreIds.push_back(coreIds[chunkIndex * laneCount + lane]);
+  return laneCoreIds;
+}
+
+} // namespace
+
+LogicalResult withScalarCoreFromBatchLane(pim::PimCoreBatchOp coreBatchOp,
+                                          unsigned lane,
+                                          llvm::function_ref<LogicalResult(pim::PimCoreOp)> callback) {
+  OwningOpRef<ModuleOp> scratchModule = ModuleOp::create(coreBatchOp.getLoc());
+  OpBuilder builder(scratchModule->getContext());
+  builder.setInsertionPointToStart(scratchModule->getBody());
+
+  size_t laneCount = static_cast<size_t>(coreBatchOp.getLaneCount());
+  size_t weightsPerLane = coreBatchOp.getWeights().size() / laneCount;
+  SmallVector<Value> laneWeights;
+  laneWeights.reserve(weightsPerLane);
+  for (size_t weightIndex = 0; weightIndex < weightsPerLane; ++weightIndex)
+    laneWeights.push_back(coreBatchOp.getWeights()[lane * weightsPerLane + weightIndex]);
+
+  auto coreIds = getBatchCoreIds(coreBatchOp);
+  auto scalarCore = pim::PimCoreOp::create(
+    builder, coreBatchOp.getLoc(), ValueRange(laneWeights), builder.getI32IntegerAttr(coreIds[lane]));
+  Block* block = builder.createBlock(&scalarCore.getBody(), scalarCore.getBody().end());
+  IRMapping mapper;
+  if (coreBatchOp.getBody().front().getNumArguments() == 1)
+    mapper.map(coreBatchOp.getBody().front().getArgument(0), coreBatchOp.getInputs()[lane]);
+
+  builder.setInsertionPointToEnd(block);
+  for (Operation& op : coreBatchOp.getBody().front()) {
+    if (isa<pim::PimHaltOp>(op)) {
+      pim::PimHaltOp::create(builder, op.getLoc());
+      continue;
+    }
+
+    if (auto sendBatchOp = dyn_cast<pim::PimSendBatchOp>(op)) {
+      pim::PimSendOp::create(builder,
+                             sendBatchOp.getLoc(),
+                             mapper.lookup(sendBatchOp.getInput()),
+                             sendBatchOp.getSizeAttr(),
+                             builder.getI32IntegerAttr(sendBatchOp.getTargetCoreIds()[lane]));
+      continue;
+    }
+
+    if (auto sendTensorBatchOp = dyn_cast<pim::PimSendTensorBatchOp>(op)) {
+      pim::PimSendTensorOp::create(
+        builder,
+        sendTensorBatchOp.getLoc(),
+        mapper.lookup(sendTensorBatchOp.getInput()),
+        builder.getDenseI32ArrayAttr(getLaneChunkCoreIds(sendTensorBatchOp.getTargetCoreIds(), laneCount, lane)));
+      continue;
+    }
+
+    if (auto receiveBatchOp = dyn_cast<pim::PimReceiveBatchOp>(op)) {
+      auto scalarReceive =
+        pim::PimReceiveOp::create(builder,
+                                  receiveBatchOp.getLoc(),
+                                  receiveBatchOp.getOutput().getType(),
+                                  mapper.lookup(receiveBatchOp.getOutputBuffer()),
+                                  receiveBatchOp.getSizeAttr(),
+                                  builder.getI32IntegerAttr(receiveBatchOp.getSourceCoreIds()[lane]));
+      mapper.map(receiveBatchOp.getOutput(), scalarReceive.getOutput());
+      continue;
+    }
+
+    if (auto receiveTensorBatchOp = dyn_cast<pim::PimReceiveTensorBatchOp>(op)) {
+      auto scalarReceive = pim::PimReceiveTensorOp::create(
+        builder,
+        receiveTensorBatchOp.getLoc(),
+        receiveTensorBatchOp.getOutput().getType(),
+        mapper.lookup(receiveTensorBatchOp.getOutputBuffer()),
+        builder.getDenseI32ArrayAttr(getLaneChunkCoreIds(receiveTensorBatchOp.getSourceCoreIds(), laneCount, lane)));
+      mapper.map(receiveTensorBatchOp.getOutput(), scalarReceive.getOutput());
+      continue;
+    }
+
+    if (auto memcpBatchOp = dyn_cast<pim::PimMemCopyHostToDevBatchOp>(op)) {
+      Value hostSource = mapper.lookupOrNull(memcpBatchOp.getHostSource());
+      if (!hostSource)
+        hostSource = memcpBatchOp.getHostSource();
+
+      auto scalarCopy = pim::PimMemCopyHostToDevOp::create(builder,
+                                                           memcpBatchOp.getLoc(),
+                                                           memcpBatchOp.getOutput().getType(),
+                                                           mapper.lookup(memcpBatchOp.getDeviceTarget()),
+                                                           hostSource,
+                                                           memcpBatchOp.getDeviceTargetOffsetAttr(),
+                                                           memcpBatchOp.getHostSourceOffsetAttr(),
+                                                           memcpBatchOp.getSizeAttr());
+      mapper.map(memcpBatchOp.getOutput(), scalarCopy.getOutput());
+      continue;
+    }
+
+    Operation* cloned = builder.clone(op, mapper);
+    for (auto [originalResult, clonedResult] : llvm::zip(op.getResults(), cloned->getResults()))
+      mapper.map(originalResult, clonedResult);
+  }
+
+  if (block->empty() || !isa<pim::PimHaltOp>(block->back()))
+    pim::PimHaltOp::create(builder, coreBatchOp.getLoc());
+  return callback(scalarCore);
+}
+
+} // namespace onnx_mlir

src/PIM/Compiler/PimBatchEmission.hpp

@@ -0,0 +1,13 @@
+#pragma once
+
+#include "llvm/ADT/STLFunctionalExtras.h"
+
+#include "src/Accelerators/PIM/Dialect/Pim/PimOps.hpp"
+
+namespace onnx_mlir {
+
+mlir::LogicalResult withScalarCoreFromBatchLane(pim::PimCoreBatchOp coreBatchOp,
+                                                unsigned lane,
+                                                llvm::function_ref<mlir::LogicalResult(pim::PimCoreOp)> callback);
+
+} // namespace onnx_mlir

src/PIM/Compiler/PimCodeGen.cpp

@@ -1,30 +1,48 @@
 #include "mlir/Dialect/Bufferization/IR/BufferizableOpInterface.h"
+#include "mlir/Dialect/Func/IR/FuncOps.h"
 #include "mlir/Dialect/MemRef/IR/MemRef.h"
+#include "mlir/IR/AsmState.h"
 #include "mlir/IR/Attributes.h"
 #include "mlir/IR/BuiltinAttributes.h"
 #include "mlir/IR/BuiltinTypes.h"
+#include "mlir/IR/Value.h"
+#include "mlir/IR/Verifier.h"
 
 #include "llvm/ADT/DenseMap.h"
-#include "llvm/ADT/SmallPtrSet.h"
+#include "llvm/ADT/StringExtras.h"
+#include "llvm/Support/Debug.h"
 #include "llvm/Support/FileSystem.h"
+#include "llvm/Support/Format.h"
 #include "llvm/Support/JSON.h"
 #include "llvm/Support/raw_ostream.h"
 
+#include <absl/types/compare.h>
 #include <algorithm>
 #include <cassert>
-#include <cmath>
+#include <cstdint>
+#include <fstream>
 #include <string>
 #include <utility>
 
 #include "Common/PimCommon.hpp"
-#include "Conversion/ONNXToSpatial/Common.hpp"
+#include "Common/IR/CompactAsmUtils.hpp"
+#include "Conversion/ONNXToSpatial/Common/Common.hpp"
+#include "src/Accelerators/PIM/Compiler/PimArtifactWriter.hpp"
+#include "src/Accelerators/PIM/Compiler/PimBatchEmission.hpp"
 #include "src/Accelerators/PIM/Compiler/PimCodeGen.hpp"
 #include "src/Accelerators/PIM/Compiler/PimCompilerOptions.hpp"
+#include "src/Accelerators/PIM/Compiler/PimWeightEmitter.hpp"
 #include "src/Accelerators/PIM/Dialect/Pim/PimOps.hpp"
 
 using namespace llvm;
 using namespace mlir;
 using namespace onnx_mlir;
+using namespace onnx_mlir::compact_asm;
+
+static size_t getValueSizeInBytes(mlir::Value value) {
+  auto type = cast<ShapedType>(value.getType());
+  return type.getNumElements() * type.getElementTypeBitWidth() / 8;
+}
 
 MemEntry* PimMemory::gatherMemEntry(mlir::Value value) {
   auto type = cast<ShapedType>(value.getType());
@@ -53,9 +71,22 @@ void PimMemory::allocateMemoryForValue(mlir::Value value, MemEntry& memEntry) {
 void PimMemory::allocateHost(ModuleOp moduleOp, func::FuncOp funcOp) {
   SmallDenseMap<memref::GlobalOp, mlir::Value, 8> globalConstants;
   SmallVector<std::pair<mlir::Value, mlir::Value>, 16> globalAliases;
+  SmallVector<mlir::Value> args;
+
+  for (mlir::Value arg : funcOp.getArguments()) {
+    gatherMemEntry(arg);
+    args.push_back(arg);
+  }
+
   funcOp.walk([&](memref::GetGlobalOp getGlobalOp) {
     if (!hasWeightAlways(getGlobalOp)) {
       auto globalMemrefOp = lookupGlobalForGetGlobal(moduleOp, getGlobalOp);
+      if (globalMemrefOp.getName().starts_with("arg")) {
+        StringRef indexStr = globalMemrefOp.getName().substr(4);
+        int index = 0;
+        llvm::to_integer(indexStr, index, 10);
+        globalAliases.push_back({getGlobalOp.getResult(), args[index]});
+      }
       auto [iter, inserted] = globalConstants.try_emplace(globalMemrefOp, getGlobalOp.getResult());
       if (inserted)
         gatherMemEntry(getGlobalOp.getResult());
@@ -64,9 +95,6 @@ void PimMemory::allocateHost(ModuleOp moduleOp, func::FuncOp funcOp) {
     }
   });
 
-  for (mlir::Value arg : funcOp.getArguments())
-    gatherMemEntry(arg);
-
   funcOp.walk([&](memref::AllocOp allocOp) {
     if (!allocOp->getParentOfType<pim::PimCoreOp>())
       gatherMemEntry(allocOp.getResult());
@@ -84,6 +112,60 @@ void PimMemory::allocateCore(Operation* op) {
   allocateGatheredMemory();
 }
 
+std::string formatMemory(uint64_t bytes) {
+  const char* units[] = {"B", "KB", "MB", "GB", "TB", "PB", "EB"};
+  int i = 0;
+  double size = static_cast<double>(bytes);
+  while (size >= 1024 && i < 6) {
+    size /= 1024;
+    i++;
+  }
+  // Formats to 2 decimal places
+  std::string out;
+  llvm::raw_string_ostream rss(out);
+  rss << llvm::format("%.2f ", size) << units[i];
+  return rss.str();
+}
+
+static void printMemoryReportRow(raw_ostream& os, const MemoryReportRow& row) {
+  os << "\tNumber of allocas: " << row.numAlloca << "\n";
+  os << "\tAllocated memory: " << formatMemory(row.sizeAlloca) << "\n";
+  os << "\tNumber of globals: " << row.numGlobal << "\n";
+  os << "\tGlobal memory: " << formatMemory(row.sizeGlobal) << "\n";
+}
+
+static MemoryReportRow addMemoryReportRows(const MemoryReportRow& lhs, const MemoryReportRow& rhs) {
+  MemoryReportRow result = lhs;
+  result.numAlloca += rhs.numAlloca;
+  result.sizeAlloca += rhs.sizeAlloca;
+  result.numGlobal += rhs.numGlobal;
+  result.sizeGlobal += rhs.sizeGlobal;
+  return result;
+}
+
+MemoryReportRow PimMemory::getReportRow() const {
+  MemoryReportRow row;
+  for (auto& [val, memEntry] : globalMemEntriesMap) {
+    if (auto op = val.getDefiningOp()) {
+      if (isa<memref::AllocOp>(op)) {
+        row.numAlloca++;
+        row.sizeAlloca += memEntry.size;
+      }
+
+      if (isa<memref::GetGlobalOp>(op)) {
+        row.numGlobal++;
+        row.sizeGlobal += memEntry.size;
+      }
+    }
+  }
+  return row;
+}
+
+void PimMemory::remove(mlir::Value val) {
+  if (auto removeIter = globalMemEntriesMap.find(val); removeIter != globalMemEntriesMap.end())
+    globalMemEntriesMap.erase(removeIter);
+}
+
 MemEntry PimMemory::getMemEntry(mlir::Value value) const {
   auto iter = globalMemEntriesMap.find(value);
   assert("Missing memEntry for value" && iter != globalMemEntriesMap.end());
@@ -124,6 +206,99 @@ size_t PimAcceleratorMemory::getValueAddress(mlir::Value value, const StaticValu
   return iter->second.address + resolvedAddress->byteOffset;
 }
 
+void PimAcceleratorMemory::reportHost() {
+  hostReportRow = hostMem.getReportRow();
+}
+
+void PimAcceleratorMemory::recordCoreReport(size_t coreId, const MemoryReportRow& row) {
+  reportEntries.push_back({MemoryReportEntry::Kind::Core, coreId, {static_cast<int32_t>(coreId)}, row});
+}
+
+void PimAcceleratorMemory::recordBatchReport(uint64_t batchId, ArrayRef<int32_t> coreIds, const MemoryReportRow& row) {
+  MemoryReportEntry entry;
+  entry.kind = MemoryReportEntry::Kind::Batch;
+  entry.id = batchId;
+  llvm::append_range(entry.coreIds, coreIds);
+  entry.row = row;
+  reportEntries.push_back(std::move(entry));
+}
+
+void PimAcceleratorMemory::flushReport() {
+  if (!fileReport.is_open())
+    return;
+
+  llvm::raw_os_ostream os(fileReport);
+  if (hostReportRow.has_value()) {
+    os << "Host:\n";
+    printMemoryReportRow(os, *hostReportRow);
+  }
+
+  if (!reportEntries.empty()) {
+    if (hostReportRow.has_value())
+      os << "\n";
+
+    llvm::stable_sort(reportEntries, [](const MemoryReportEntry& lhs, const MemoryReportEntry& rhs) {
+      if (lhs.kind != rhs.kind)
+        return lhs.kind == MemoryReportEntry::Kind::Batch;
+
+      const MemoryReportRow& lhsRow = lhs.row;
+      const MemoryReportRow& rhsRow = rhs.row;
+      if (lhsRow.sizeAlloca != rhsRow.sizeAlloca)
+        return lhsRow.sizeAlloca > rhsRow.sizeAlloca;
+      if (lhsRow.numAlloca != rhsRow.numAlloca)
+        return lhsRow.numAlloca > rhsRow.numAlloca;
+      if (lhsRow.sizeGlobal != rhsRow.sizeGlobal)
+        return lhsRow.sizeGlobal > rhsRow.sizeGlobal;
+      if (lhsRow.numGlobal != rhsRow.numGlobal)
+        return lhsRow.numGlobal > rhsRow.numGlobal;
+      return lhs.id < rhs.id;
+    });
+
+    for (size_t index = 0; index < reportEntries.size();) {
+      size_t runEnd = index + 1;
+      while (runEnd < reportEntries.size() && reportEntries[runEnd].kind == reportEntries[index].kind
+             && reportEntries[runEnd].row == reportEntries[index].row) {
+        ++runEnd;
+      }
+
+      if (reportEntries[index].kind == MemoryReportEntry::Kind::Batch) {
+        os << "Batch ";
+        for (size_t batchIndex = index; batchIndex < runEnd; ++batchIndex) {
+          if (batchIndex != index)
+            os << ",\n      ";
+          os << reportEntries[batchIndex].id << " (cores ";
+          printCompressedIntegerEntries(os, ArrayRef<int32_t>(reportEntries[batchIndex].coreIds));
+          os << ")";
+        }
+      }
+      else {
+        llvm::SmallVector<int32_t, 8> coreIds;
+        for (size_t coreIndex = index; coreIndex < runEnd; ++coreIndex)
+          coreIds.push_back(reportEntries[coreIndex].coreIds.front());
+        os << "Core ";
+        printCompressedIntegerEntries(os, ArrayRef<int32_t>(coreIds));
+      }
+      os << ":\n";
+      printMemoryReportRow(os, reportEntries[index].row);
+      if (runEnd < reportEntries.size())
+        os << "\n";
+
+      index = runEnd;
+    }
+  }
+
+  os.flush();
+  fileReport.close();
+}
+
+void PimAcceleratorMemory::clean(mlir::Operation* op) {
+  for (auto value : op->getResults()) {
+    hostMem.remove(value);
+    for (auto& device : deviceMem)
+      device.second.remove(value);
+  }
+}
+
 json::Object PimCodeGen::createEmptyOffset() {
   json::Object offset;
   offset["offset_select"] = 0;
@@ -131,6 +306,12 @@ json::Object PimCodeGen::createEmptyOffset() {
   return offset;
 }
 
+size_t PimCodeGen::remapCoreId(size_t coreId) const {
+  auto it = emittedCoreIds.find(coreId);
+  assert(it != emittedCoreIds.end() && "Missing emitted core id remapping");
+  return it->second;
+}
+
 static json::Object createRs1OnlyOffset() {
   json::Object offset;
   offset["offset_select"] = 1;
@@ -190,7 +371,7 @@ void PimCodeGen::emitCommunicationOp(StringRef opName, size_t bufferAddr, size_t
   json::Object json;
   json["op"] = opName;
   json["rd"] = 0;
-  json["core"] = coreId;
+  json["core"] = remapCoreId(coreId);
   json["size"] = size;
   json["offset"] = createEmptyOffset();
   emitInstruction(std::move(json));
@@ -242,10 +423,62 @@ void PimCodeGen::codeGenReceiveOp(pim::PimReceiveOp receiveOp, const StaticValue
     "recv", addressOf(receiveOp.getOutputBuffer(), knowledge), receiveOp.getSourceCoreId(), receiveOp.getSize());
 }
 
+void PimCodeGen::codeGenReceiveTensorOp(pim::PimReceiveTensorOp receiveTensorOp,
+                                        const StaticValueKnowledge& knowledge) const {
+  size_t outputAddr = addressOf(receiveTensorOp.getOutputBuffer(), knowledge);
+  size_t chunkSize = getValueSizeInBytes(receiveTensorOp.getOutputBuffer()) / receiveTensorOp.getSourceCoreIds().size();
+  for (auto [chunkIndex, sourceCoreId] : llvm::enumerate(receiveTensorOp.getSourceCoreIds()))
+    emitCommunicationOp("recv", outputAddr + chunkIndex * chunkSize, sourceCoreId, chunkSize);
+}
+
 void PimCodeGen::codeGenSendOp(pim::PimSendOp sendOp, const StaticValueKnowledge& knowledge) const {
   emitCommunicationOp("send", addressOf(sendOp.getInput(), knowledge), sendOp.getTargetCoreId(), sendOp.getSize());
 }
 
+void PimCodeGen::codeGenSendTensorOp(pim::PimSendTensorOp sendTensorOp, const StaticValueKnowledge& knowledge) const {
+  size_t inputAddr = addressOf(sendTensorOp.getInput(), knowledge);
+  size_t chunkSize = getValueSizeInBytes(sendTensorOp.getInput()) / sendTensorOp.getTargetCoreIds().size();
+  for (auto [chunkIndex, targetCoreId] : llvm::enumerate(sendTensorOp.getTargetCoreIds()))
+    emitCommunicationOp("send", inputAddr + chunkIndex * chunkSize, targetCoreId, chunkSize);
+}
+
+void PimCodeGen::codeGenConcatOp(pim::PimConcatOp concatOp, const StaticValueKnowledge& knowledge) const {
+  auto outputType = cast<ShapedType>(concatOp.getOutputBuffer().getType());
+  assert(outputType.hasStaticShape() && "concat codegen requires static output shape");
+
+  int64_t axis = concatOp.getAxis();
+  ArrayRef<int64_t> outputShape = outputType.getShape();
+  size_t elementSize = outputType.getElementTypeBitWidth() / 8;
+  size_t outputAddr = addressOf(concatOp.getOutputBuffer(), knowledge);
+
+  size_t outerCount = 1;
+  for (int64_t dim = 0; dim < axis; ++dim)
+    outerCount *= static_cast<size_t>(outputShape[dim]);
+
+  size_t innerCount = 1;
+  for (size_t dim = static_cast<size_t>(axis) + 1; dim < outputShape.size(); ++dim)
+    innerCount *= static_cast<size_t>(outputShape[dim]);
+
+  size_t outputConcatDim = static_cast<size_t>(outputShape[axis]);
+  size_t concatOffset = 0;
+  for (mlir::Value input : concatOp.getInputs()) {
+    auto inputType = cast<ShapedType>(input.getType());
+    assert(inputType.hasStaticShape() && "concat codegen requires static input shapes");
+
+    size_t inputConcatDim = static_cast<size_t>(inputType.getDimSize(axis));
+    size_t blockSizeInBytes = inputConcatDim * innerCount * elementSize;
+    size_t inputAddr = addressOf(input, knowledge);
+
+    for (size_t outerIndex = 0; outerIndex < outerCount; ++outerIndex) {
+      size_t dstOffset = (outerIndex * outputConcatDim + concatOffset) * innerCount * elementSize;
+      size_t srcOffset = outerIndex * inputConcatDim * innerCount * elementSize;
+      emitMemCopyOp("lmv", outputAddr, dstOffset, inputAddr, srcOffset, blockSizeInBytes, "len");
+    }
+
+    concatOffset += inputConcatDim;
+  }
+}
+
 template <typename MVMTy>
 void PimCodeGen::codeGenMVMLikeOp(size_t mvmId,
                                   MVMTy mvmLikeOp,
@@ -256,11 +489,6 @@ void PimCodeGen::codeGenMVMLikeOp(size_t mvmId,
   // TODO: save weights somewhere (if transposeMatrix=true, transpose the weight matrix)
 }
 
-static size_t getValueSizeInBytes(mlir::Value value) {
-  auto type = cast<ShapedType>(value.getType());
-  return type.getNumElements() * type.getElementTypeBitWidth() / 8;
-}
-
 void PimCodeGen::codeGenVVAddOp(pim::PimVVAddOp vvaddOp, const StaticValueKnowledge& knowledge) const {
   auto outputBufferAddr = addressOf(vvaddOp.getOutputBuffer(), knowledge);
   auto lhsAddr = addressOf(vvaddOp.getLhs(), knowledge);
@@ -412,6 +640,8 @@ void PimCodeGen::codeGenVSoftmaxOp(pim::PimVSoftmaxOp vsoftmaxOp, const StaticVa
   emitInstruction(std::move(json));
 }
 
+void PimCodeGen::codeGetGlobalOp(memref::GetGlobalOp getGlobalOp, const StaticValueKnowledge& knowledge) const {}
+
 void PimCodeGen::codeGenTransposeOp(pim::PimTransposeOp transposeOp, const StaticValueKnowledge& knowledge) const {
   auto srcAddr = addressOf(transposeOp.getInput(), knowledge);
   auto dstAddr = addressOf(transposeOp.getOutputBuffer(), knowledge);
@@ -474,67 +704,59 @@ std::string getMemorySizeAsString(size_t size) {
   return std::to_string(size) + " Bytes";
 }
 
-static SmallVector<unsigned, 8> getUsedWeightIndices(pim::PimCoreOp coreOp) {
+static SmallVector<unsigned, 8> getUsedWeightIndices(Block& block) {
   SmallVector<unsigned, 8> indices;
   auto addIndex = [&](unsigned weightIndex) {
     if (!llvm::is_contained(indices, weightIndex))
       indices.push_back(weightIndex);
   };
 
-  coreOp.walk([&](pim::PimMVMOp mvmOp) { addIndex(mvmOp.getWeightIndex()); });
+  block.walk([&](pim::PimVMMOp vmmOp) { addIndex(vmmOp.getWeightIndex()); });
-  coreOp.walk([&](pim::PimVMMOp vmmOp) { addIndex(vmmOp.getWeightIndex()); });
   llvm::sort(indices);
   return indices;
 }
 
-/// Write global constant data into a binary memory image at their allocated addresses.
+static SmallVector<unsigned, 8> getUsedWeightIndices(pim::PimCoreOp coreOp) {
-static OnnxMlirCompilerErrorCodes
+  return getUsedWeightIndices(coreOp.getBody().front());
-writeMemoryBinary(ModuleOp moduleOp, func::FuncOp funcOp, PimAcceleratorMemory& memory, StringRef outputDirPath) {
-  auto memoryFilePath = (outputDirPath + "/memory.bin").str();
-  std::error_code errorCode;
-  raw_fd_ostream memoryFileStream(memoryFilePath, errorCode, sys::fs::OF_None);
-  if (errorCode) {
-    errs() << "Error while opening memory file " << memoryFilePath << ": " << errorCode.message() << '\n';
-    return InvalidOutputFileAccess;
 }
 
-  std::vector<char> memoryBuffer(memory.hostMem.getFirstAvailableAddress(), 0);
+static SmallVector<int32_t> getBatchCoreIds(pim::PimCoreBatchOp coreBatchOp) {
-
+  auto coreIdsAttr = coreBatchOp->getAttrOfType<DenseI32ArrayAttr>(onnx_mlir::kCoreIdsAttrName);
-  SmallPtrSet<Operation*, 16> writtenGlobals;
+  assert(coreIdsAttr && "pim.core_batch requires coreIds array attribute");
-  funcOp.walk([&](memref::GetGlobalOp getGlobalOp) {
+  return SmallVector<int32_t>(coreIdsAttr.asArrayRef().begin(), coreIdsAttr.asArrayRef().end());
-    if (hasWeightAlways(getGlobalOp))
-      return;
-    auto globalOp = lookupGlobalForGetGlobal(moduleOp, getGlobalOp);
-    if (!globalOp)
-      return;
-    if (!writtenGlobals.insert(globalOp.getOperation()).second)
-      return;
-    auto initialValue = globalOp.getInitialValue();
-    if (!initialValue)
-      return;
-    auto denseAttr = dyn_cast<DenseElementsAttr>(*initialValue);
-    if (!denseAttr)
-      return;
-
-    MemEntry memEntry = memory.hostMem.getMemEntry(getGlobalOp.getResult());
-    ArrayRef<char> rawData = denseAttr.getRawData();
-    char* dst = memoryBuffer.data() + memEntry.address;
-
-    if (denseAttr.isSplat()) {
-      size_t elementSize = rawData.size();
-      assert(elementSize * getGlobalOp.getType().getNumElements() == memEntry.size && "Data size mismatch");
-      for (size_t offset = 0; offset < memEntry.size; offset += elementSize)
-        std::memcpy(dst + offset, rawData.data(), std::min(elementSize, memEntry.size - offset));
 }
-    else {
+
-      assert(rawData.size() == memEntry.size && "Data size mismatch");
+static SmallVector<Operation*> collectTopLevelCoreLikeOps(func::FuncOp funcOp) {
-      std::memcpy(dst, rawData.data(), rawData.size());
+  SmallVector<Operation*> coreLikeOps;
+  for (Operation& op : funcOp.getBody().front())
+    if (dyn_cast<pim::PimCoreOp>(&op) || dyn_cast<pim::PimCoreBatchOp>(&op))
+      coreLikeOps.push_back(&op);
+  return coreLikeOps;
 }
+
+static void aliasMaterializedHostGlobals(ModuleOp moduleOp,
+                                         func::FuncOp funcOp,
+                                         pim::PimCoreOp coreOp,
+                                         PimAcceleratorMemory& memory) {
+  coreOp.walk([&](memref::GetGlobalOp getGlobalOp) {
+    if (hasWeightAlways(getGlobalOp) || memory.memEntriesMap.contains(getGlobalOp.getResult()))
+      return;
+
+    auto targetGlobal = lookupGlobalForGetGlobal(moduleOp, getGlobalOp);
+    if (!targetGlobal)
+      return;
+
+    mlir::Value aliasedValue;
+    funcOp.walk([&](memref::GetGlobalOp candidate) {
+      if (aliasedValue || candidate == getGlobalOp || !memory.memEntriesMap.contains(candidate.getResult()))
+        return;
+      if (lookupGlobalForGetGlobal(moduleOp, candidate) == targetGlobal)
+        aliasedValue = candidate.getResult();
     });
 
-  memoryFileStream.write(memoryBuffer.data(), memoryBuffer.size());
+    if (aliasedValue)
-  memoryFileStream.close();
+      memory.memEntriesMap[getGlobalOp.getResult()] = memory.memEntriesMap[aliasedValue];
-  return CompilerSuccess;
+  });
 }
 
 /// Dispatch all operations in a core region to the appropriate code generator.
@@ -553,12 +775,16 @@ static int64_t codeGenCoreOps(Block& block, PimCodeGen& coreCodeGen) {
         coreCodeGen.codeGenLmvOp(lmvOp, knowledge);
       else if (auto receiveOp = dyn_cast<pim::PimReceiveOp>(op))
         coreCodeGen.codeGenReceiveOp(receiveOp, knowledge);
+      else if (auto receiveTensorOp = dyn_cast<pim::PimReceiveTensorOp>(op))
+        coreCodeGen.codeGenReceiveTensorOp(receiveTensorOp, knowledge);
       else if (auto sendOp = dyn_cast<pim::PimSendOp>(op))
         coreCodeGen.codeGenSendOp(sendOp, knowledge);
+      else if (auto sendTensorOp = dyn_cast<pim::PimSendTensorOp>(op))
+        coreCodeGen.codeGenSendTensorOp(sendTensorOp, knowledge);
+      else if (auto concatOp = dyn_cast<pim::PimConcatOp>(op))
+        coreCodeGen.codeGenConcatOp(concatOp, knowledge);
       else if (auto vmmOp = dyn_cast<pim::PimVMMOp>(op))
         coreCodeGen.codeGenMVMLikeOp<pim::PimVMMOp>(vmmOp.getWeightIndex(), vmmOp, true, knowledge);
-      else if (auto mvmOp = dyn_cast<pim::PimMVMOp>(op))
-        coreCodeGen.codeGenMVMLikeOp<pim::PimMVMOp>(mvmOp.getWeightIndex(), mvmOp, false, knowledge);
       else if (auto transposeOp = dyn_cast<pim::PimTransposeOp>(op))
         coreCodeGen.codeGenTransposeOp(transposeOp, knowledge);
       else if (auto vvaddOp = dyn_cast<pim::PimVVAddOp>(op))
@@ -581,9 +807,10 @@ static int64_t codeGenCoreOps(Block& block, PimCodeGen& coreCodeGen) {
         coreCodeGen.codeGenVSigmOp(vsigmOp, knowledge);
       else if (auto vsoftmaxOp = dyn_cast<pim::PimVSoftmaxOp>(op))
         coreCodeGen.codeGenVSoftmaxOp(vsoftmaxOp, knowledge);
+      else if (auto getGlobalOp = dyn_cast<memref::GetGlobalOp>(op))
+        coreCodeGen.codeGetGlobalOp(getGlobalOp, knowledge);
       else {
         op.emitError("Unsupported codegen for this operation");
-        op.dump();
         return failure();
       }
       processedOperations++;
@@ -592,225 +819,6 @@ static int64_t codeGenCoreOps(Block& block, PimCodeGen& coreCodeGen) {
   return failed(result) ? -1 : static_cast<int64_t>(processedOperations);
 }
 
-/// Write crossbar weight matrices as padded binary files for a single core.
-static OnnxMlirCompilerErrorCodes writeCrossbarWeights(ModuleOp moduleOp,
-                                                       pim::PimCoreOp coreOp,
-                                                       StringRef coreWeightsDirPath,
-                                                       json::Array& xbarsPerGroup) {
-  int64_t xbarSize = crossbarSize.getValue();
-  std::error_code errorCode;
-  size_t weightIndex = 0;
-
-  for (auto weight : coreOp.getWeights()) {
-    xbarsPerGroup.push_back(weightIndex);
-
-    auto getGlobalOp = weight.getDefiningOp<memref::GetGlobalOp>();
-    if (!getGlobalOp) {
-      coreOp.emitWarning("Weight is not from a memref.get_global at index " + std::to_string(weightIndex));
-      weightIndex++;
-      continue;
-    }
-
-    auto globalOp = lookupGlobalForGetGlobal(moduleOp, getGlobalOp);
-    if (!globalOp) {
-      coreOp.emitWarning("Could not find memref.global for weight at index " + std::to_string(weightIndex));
-      weightIndex++;
-      continue;
-    }
-
-    auto initialValue = globalOp.getInitialValue();
-    if (!initialValue) {
-      coreOp.emitWarning("memref.global has no initial value at index " + std::to_string(weightIndex));
-      weightIndex++;
-      continue;
-    }
-
-    auto denseAttr = dyn_cast<DenseElementsAttr>(*initialValue);
-    if (!denseAttr) {
-      coreOp.emitWarning("memref.global initial value is not dense at index " + std::to_string(weightIndex));
-      weightIndex++;
-      continue;
-    }
-
-    auto type = denseAttr.getType();
-    auto shape = type.getShape();
-    assert(isMatrixShape(shape) && "Weight matrix must be 2-dimensional");
-    int64_t numRows = shape[0];
-    int64_t numCols = shape[1];
-    assert(numRows <= xbarSize && numCols <= xbarSize && "Weight dimensions must not exceed crossbar size");
-
-    size_t elementByteWidth = type.getElementType().getIntOrFloatBitWidth() / 8;
-
-    auto weightFilePath = (coreWeightsDirPath + "/crossbar_" + std::to_string(weightIndex) + ".bin").str();
-    raw_fd_ostream weightFileStream(weightFilePath, errorCode, sys::fs::OF_None);
-    if (errorCode) {
-      errs() << "Error while opening weight file `" << weightFilePath << "`: " << errorCode.message() << '\n';
-      return InvalidOutputFileAccess;
-    }
-
-    uint64_t zero = 0;
-    for (int64_t row = 0; row < xbarSize; row++) {
-      for (int64_t col = 0; col < xbarSize; col++) {
-        if (row < numRows && col < numCols) {
-          int64_t index = row * numCols + col;
-          APInt bits = denseAttr.getValues<APFloat>()[index].bitcastToAPInt();
-          uint64_t word = bits.getZExtValue();
-          weightFileStream.write(reinterpret_cast<const char*>(&word), elementByteWidth);
-        }
-        else {
-          weightFileStream.write(reinterpret_cast<const char*>(&zero), elementByteWidth);
-        }
-      }
-    }
-
-    weightFileStream.close();
-    weightIndex++;
-  }
-
-  return CompilerSuccess;
-}
-
-llvm::DenseMap<pim::PimCoreOp, llvm::DenseMap<mlir::Value, std::string>>
-createAndPopulateWeightFolder(func::FuncOp funcOp, StringRef outputDirPath) {
-  ModuleOp moduleOp = funcOp->getParentOfType<ModuleOp>();
-  auto coreWeightsDirPath = outputDirPath + "/weights";
-  auto error = sys::fs::create_directory(coreWeightsDirPath);
-  assert(!error && "Error creating weights directory");
-  size_t indexFileName = 0;
-
-  int64_t xbarSize = crossbarSize.getValue();
-  llvm::DenseMap<pim::PimCoreOp, llvm::DenseMap<mlir::Value, std::string>> mapCoreWeightToFileName;
-  llvm::DenseMap<memref::GlobalOp, std::string> mapGlobalOpToFileName;
-
-  for (pim::PimCoreOp coreOp : funcOp.getOps<pim::PimCoreOp>()) {
-    for (unsigned index : getUsedWeightIndices(coreOp)) {
-      if (index >= coreOp.getWeights().size()) {
-        coreOp.emitWarning("Weight index " + std::to_string(index) + " is out of range");
-        assert(index < coreOp.getWeights().size() && "Weight index is out of range");
-      }
-      mlir::Value weight = coreOp.getWeights()[index];
-
-      auto getGlobalOp = weight.getDefiningOp<memref::GetGlobalOp>();
-      if (!getGlobalOp) {
-        coreOp.emitWarning("Weight is not from a memref.get_global at index " + std::to_string(index));
-        assert(!getGlobalOp && "Weight is not from a memref.get_global");
-      }
-
-      auto globalOp = lookupGlobalForGetGlobal(moduleOp, getGlobalOp);
-      if (!globalOp) {
-        coreOp.emitWarning("Could not find memref.global for weight at index " + std::to_string(index));
-        assert(!globalOp && "Could not find memref.global");
-      }
-
-      auto initialValue = globalOp.getInitialValue();
-      if (!initialValue) {
-        coreOp.emitWarning("memref.global has no initial value at index " + std::to_string(index));
-        assert(!initialValue && "memref.global has no initial value");
-      }
-
-      auto denseAttr = dyn_cast<DenseElementsAttr>(*initialValue);
-      if (!denseAttr) {
-        coreOp.emitWarning("memref.global initial value is not dense at index " + std::to_string(index));
-        assert(!denseAttr && "memref.global initial value is not dense");
-      }
-
-      if (mapGlobalOpToFileName.contains(globalOp)) {
-        auto& fileName = mapGlobalOpToFileName[globalOp];
-        std::pair<mlir::Value, std::string> weightToFile = {weight, fileName};
-        mapCoreWeightToFileName[coreOp].insert(weightToFile);
-        continue;
-      }
-
-      auto type = denseAttr.getType();
-      auto shape = type.getShape();
-      assert(isMatrixShape(shape) && "Weight matrix must be 2-dimensional");
-      int64_t numRows = shape[0];
-      int64_t numCols = shape[1];
-      assert(numRows <= xbarSize && numCols <= xbarSize && "Weight dimensions must not exceed crossbar size");
-
-      size_t elementByteWidth = type.getElementType().getIntOrFloatBitWidth() / 8;
-
-      std::string newFileName = "crossbar_" + std::to_string(indexFileName++) + ".bin";
-      auto weightFilePath = (coreWeightsDirPath + "/" + newFileName).str();
-      std::error_code errorCode;
-      raw_fd_ostream weightFileStream(weightFilePath, errorCode, sys::fs::OF_None);
-      if (errorCode) {
-        errs() << "Error while opening weight file `" << weightFilePath << "`: " << errorCode.message() << '\n';
-        assert(errorCode);
-      }
-
-      uint64_t zero = 0;
-      for (int64_t row = 0; row < xbarSize; row++) {
-        for (int64_t col = 0; col < xbarSize; col++) {
-          if (row < numRows && col < numCols) {
-            int64_t index = row * numCols + col;
-            APInt bits = denseAttr.getValues<APFloat>()[index].bitcastToAPInt();
-            uint64_t word = bits.getZExtValue();
-            weightFileStream.write(reinterpret_cast<const char*>(&word), elementByteWidth);
-          }
-          else {
-            weightFileStream.write(reinterpret_cast<const char*>(&zero), elementByteWidth);
-          }
-        }
-      }
-
-      weightFileStream.close();
-      mapGlobalOpToFileName.insert({globalOp, newFileName});
-      mapCoreWeightToFileName[coreOp].insert({weight, newFileName});
-    }
-  }
-  return mapCoreWeightToFileName;
-}
-
-/// Write the top-level PIM configuration JSON (core count, crossbar config, I/O addresses).
-static OnnxMlirCompilerErrorCodes writeConfigJson(func::FuncOp funcOp,
-                                                  PimAcceleratorMemory& memory,
-                                                  size_t coreCount,
-                                                  json::Object xbarsPerArrayGroup,
-                                                  StringRef outputDirPath) {
-  json::Object configJson;
-
-  // +1 because pimsim-nn also considers the host as a core
-  configJson["core_cnt"] = coreCount + 1;
-
-  // TODO: Should this be based on the floating point type used in the model?
-  // The 2 following values determine the bitwidth of the vectors' elements: bitwidth = adc_count * cell_precision
-
-  // Number of ADC for MVM units
-  configJson["adc_count"] = 16;
-  // The bit precision of each ADC
-  configJson["cell_precision"] = 2;
-
-  // Crossbar configuration
-  configJson["xbar_array_count"] = crossbarCountInCore.getValue();
-  configJson["xbar_size"] = {crossbarSize.getValue(), crossbarSize.getValue()};
-  configJson["array_group_map"] = std::move(xbarsPerArrayGroup);
-
-  // Memory layout of inputs and outputs
-  json::Array inputsAddresses;
-  for (BlockArgument input : funcOp.getArguments())
-    inputsAddresses.push_back(memory.getValueAddress(input));
-  configJson["inputs_addresses"] = std::move(inputsAddresses);
-
-  json::Array outputsAddresses;
-  for (func::ReturnOp returnOp : funcOp.getOps<func::ReturnOp>())
-    for (mlir::Value output : returnOp.getOperands())
-      outputsAddresses.push_back(memory.getValueAddress(output));
-  configJson["outputs_addresses"] = std::move(outputsAddresses);
-
-  auto configPath = (outputDirPath + "/config.json").str();
-  std::error_code errorCode;
-  raw_fd_ostream jsonOS(configPath, errorCode);
-  if (errorCode) {
-    errs() << "Error while opening config file: " << errorCode.message() << '\n';
-    return InvalidOutputFileAccess;
-  }
-  jsonOS << json::Value(std::move(configJson)) << '\n';
-  jsonOS.close();
-
-  return CompilerSuccess;
-}
-
 OnnxMlirCompilerErrorCodes onnx_mlir::compileToPimJson(ModuleOp& moduleOp, std::string& outputDirPath) {
   if (!outputDirPath.empty()) {
     if (auto error = sys::fs::create_directory(outputDirPath)) {
@@ -826,33 +834,49 @@ OnnxMlirCompilerErrorCodes onnx_mlir::compileToPimJson(ModuleOp& moduleOp, std::
 
   PimAcceleratorMemory memory;
   memory.hostMem.allocateHost(moduleOp, funcOp);
+  memory.reportHost();
 
   if (auto err = writeMemoryBinary(moduleOp, funcOp, memory, outputDirPath))
     return err;
 
-  // Write empty host core file
+  if (auto err = writeHostCoreJson(outputDirPath))
-  std::error_code errorCode;
+    return err;
-  auto outputHostCorePath = outputDirPath + "/core_0.json";
-  raw_fd_ostream hostFileStream(outputHostCorePath, errorCode);
-  if (errorCode) {
-    errs() << "Error while opening host core file `" << outputHostCorePath << "`: " << errorCode.message() << '\n';
-    return InvalidOutputFileAccess;
-  }
-  // The host core json contains 2 random instructions, just to make pimsim-nn happy
-  hostFileStream << "[{\"imm\":0,\"op\":\"sldi\",\"rd\":0},{\"imm\":0,\"op\":\"sldi\",\"rd\":0}]";
-  hostFileStream.close();
 
   // For each core, specify the number of crossbar per array group.
   // This implementation always assigns one crossbar per group.
   json::Object xbarsPerArrayGroup;
-  size_t coreCount = 0;
+  size_t maxCoreId = 0;
+  uint64_t nextBatchReportId = 0;
 
   // Create Weight Folder
   auto mapCoreWeightToFileName = createAndPopulateWeightFolder(funcOp, outputDirPath);
 
-  for (auto coreOp : funcOp.getOps<pim::PimCoreOp>()) {
+  SmallVector<Operation*> coreLikeOps = collectTopLevelCoreLikeOps(funcOp);
-    auto coreId = coreOp.getCoreId();
+  llvm::DenseMap<size_t, size_t> emittedCoreIds;
-    coreCount++;
+  size_t nextEmittedCoreId = 1;
+
+  for (Operation* op : coreLikeOps) {
+    if (auto coreOp = dyn_cast<pim::PimCoreOp>(op)) {
+      size_t originalCoreId = static_cast<size_t>(coreOp.getCoreId());
+      if (!emittedCoreIds.contains(originalCoreId))
+        emittedCoreIds[originalCoreId] = nextEmittedCoreId++;
+      continue;
+    }
+
+    auto coreBatchOp = cast<pim::PimCoreBatchOp>(op);
+    auto batchCoreIds = getBatchCoreIds(coreBatchOp);
+    for (unsigned lane = 0; lane < static_cast<unsigned>(coreBatchOp.getLaneCount()); ++lane) {
+      size_t originalCoreId = static_cast<size_t>(batchCoreIds[lane]);
+      if (!emittedCoreIds.contains(originalCoreId))
+        emittedCoreIds[originalCoreId] = nextEmittedCoreId++;
+    }
+  }
+
+  for (Operation* op : coreLikeOps) {
+    auto emitCore = [&](pim::PimCoreOp coreOp, bool temporaryCore) -> OnnxMlirCompilerErrorCodes {
+      size_t originalCoreId = static_cast<size_t>(coreOp.getCoreId());
+      size_t coreId = emittedCoreIds.lookup(originalCoreId);
+      maxCoreId = std::max(maxCoreId, coreId);
 
       std::error_code errorCode;
       auto outputCorePath = outputDirPath + "/core_" + std::to_string(coreId) + ".json";
@@ -863,7 +887,8 @@ OnnxMlirCompilerErrorCodes onnx_mlir::compileToPimJson(ModuleOp& moduleOp, std::
       }
       coreFileStream << '[';
 
-    PimCodeGen coreCodeGen(memory, coreFileStream);
+      PimCodeGen coreCodeGen(memory, coreFileStream, emittedCoreIds);
+      aliasMaterializedHostGlobals(moduleOp, funcOp, coreOp, memory);
       memory.getOrCreateDeviceMem(coreId).allocateCore(coreOp);
 
       int64_t processedOperations = codeGenCoreOps(coreOp.getBody().front(), coreCodeGen);
@@ -871,19 +896,17 @@ OnnxMlirCompilerErrorCodes onnx_mlir::compileToPimJson(ModuleOp& moduleOp, std::
         return CompilerFailure;
       assert(processedOperations > 0);
 
-    // Remove trailing comma, close JSON array
       coreFileStream.seek(coreFileStream.tell() - 1);
       coreFileStream << ']';
       coreFileStream.close();
 
-    // Write crossbar weights for this core
       auto coreWeightsDirPath = outputDirPath + "/core_" + std::to_string(coreId);
       if (auto error = sys::fs::create_directory(coreWeightsDirPath)) {
         errs() << "Error creating core directory: " << coreWeightsDirPath << ": " << error.message() << '\n';
         return InvalidOutputFileAccess;
       }
 
-    auto& mapWeightToFile = mapCoreWeightToFileName[coreOp];
+      auto& mapWeightToFile = mapCoreWeightToFileName[originalCoreId];
       json::Array xbarsPerGroup;
       for (unsigned index : getUsedWeightIndices(coreOp)) {
         if (index >= coreOp.getWeights().size()) {
@@ -897,14 +920,48 @@ OnnxMlirCompilerErrorCodes onnx_mlir::compileToPimJson(ModuleOp& moduleOp, std::
         if (auto error = sys::fs::create_link(outputDirPath + "/weights/" + fileName,
                                               coreWeightsDirPath + "/crossbar_" + std::to_string(index) + ".bin")) {
           errs() << "Error creating link file: " << (outputDirPath + "/weights/" + fileName) << " to "
-               << (coreWeightsDirPath + "/crossbar_" + std::to_string(index) + ".bin") << "\nError:" << error.message()
+                 << (coreWeightsDirPath + "/crossbar_" + std::to_string(index) + ".bin")
-               << '\n';
+                 << "\nError:" << error.message() << '\n';
           return InvalidOutputFileAccess;
         }
       }
 
       xbarsPerArrayGroup["core" + std::to_string(coreId)] = std::move(xbarsPerGroup);
+      if (temporaryCore)
+        coreOp.walk([&memory](Operation* op) { memory.clean(op); });
+      return CompilerSuccess;
+    };
+
+    if (auto coreOp = dyn_cast<pim::PimCoreOp>(op)) {
+      if (auto err = emitCore(coreOp, false))
+        return err;
+      memory.recordCoreReport(emittedCoreIds.lookup(static_cast<size_t>(coreOp.getCoreId())),
+                              memory.getOrCreateDeviceMem(emittedCoreIds.lookup(static_cast<size_t>(coreOp.getCoreId())))
+                                .getReportRow());
+      continue;
     }
 
-  return writeConfigJson(funcOp, memory, coreCount, std::move(xbarsPerArrayGroup), outputDirPath);
+    auto coreBatchOp = cast<pim::PimCoreBatchOp>(op);
+    auto batchCoreIds = getBatchCoreIds(coreBatchOp);
+    SmallVector<int32_t> reportedCoreIds;
+    reportedCoreIds.reserve(batchCoreIds.size());
+    MemoryReportRow batchRow;
+    for (unsigned lane = 0; lane < static_cast<unsigned>(coreBatchOp.getLaneCount()); ++lane) {
+      OnnxMlirCompilerErrorCodes laneResult = CompilerSuccess;
+      if (failed(withScalarCoreFromBatchLane(coreBatchOp, lane, [&](pim::PimCoreOp coreOp) {
+            size_t originalCoreId = static_cast<size_t>(batchCoreIds[lane]);
+            size_t coreId = emittedCoreIds.lookup(originalCoreId);
+            reportedCoreIds.push_back(static_cast<int32_t>(coreId));
+            laneResult = emitCore(coreOp, true);
+            if (laneResult == CompilerSuccess)
+              batchRow = addMemoryReportRows(batchRow, memory.getOrCreateDeviceMem(coreId).getReportRow());
+            return laneResult == CompilerSuccess ? success() : failure();
+          })))
+        return laneResult == CompilerSuccess ? CompilerFailure : laneResult;
+    }
+    memory.recordBatchReport(nextBatchReportId++, reportedCoreIds, batchRow);
+  }
+
+  memory.flushReport();
+  return writeConfigJson(funcOp, memory, maxCoreId, std::move(xbarsPerArrayGroup), outputDirPath);
 }

src/PIM/Compiler/PimCodeGen.hpp

@@ -1,7 +1,14 @@
 #pragma once
 
+#include "mlir/IR/Operation.h"
+
 #include "llvm-project/clang/include/clang/Basic/LLVM.h"
+#include "llvm/ADT/DenseMap.h"
 #include "llvm/Support/JSON.h"
+#include "llvm/Support/raw_os_ostream.h"
+
+#include <fstream>
+#include <optional>
 
 #include "onnx-mlir/Compiler/OMCompilerTypes.h"
 #include "src/Accelerators/PIM/Common/PimCommon.hpp"
@@ -14,12 +21,34 @@ struct MemEntry {
   size_t size;
 };
 
+struct MemoryReportRow {
+  uint64_t numAlloca = 0;
+  uint64_t sizeAlloca = 0;
+  uint64_t numGlobal = 0;
+  uint64_t sizeGlobal = 0;
+
+  bool operator==(const MemoryReportRow& other) const {
+    return numAlloca == other.numAlloca && sizeAlloca == other.sizeAlloca && numGlobal == other.numGlobal
+           && sizeGlobal == other.sizeGlobal;
+  }
+};
+
+struct MemoryReportEntry {
+  enum class Kind {
+    Core,
+    Batch
+  };
+
+  Kind kind = Kind::Core;
+  uint64_t id = 0;
+  llvm::SmallVector<int32_t, 8> coreIds;
+  MemoryReportRow row;
+};
+
 class PimMemory {
   llvm::SmallVector<std::pair<MemEntry, mlir::Value>, 32> memEntries;
   llvm::SmallDenseMap<mlir::Value, MemEntry, 32>& globalMemEntriesMap;
 
-  size_t maxSize = 0; // 0 for unbounded memory
-  size_t startAddress = 0;
   size_t minAlignment = 4;
   size_t firstAvailableAddress = 0;
 
@@ -33,6 +62,8 @@ public:
 
   void allocateHost(mlir::ModuleOp moduleOp, mlir::func::FuncOp funcOp);
   void allocateCore(mlir::Operation* op);
+  MemoryReportRow getReportRow() const;
+  void remove(mlir::Value val);
 
   size_t getFirstAvailableAddress() const { return firstAvailableAddress; }
   MemEntry getMemEntry(mlir::Value value) const;
@@ -45,23 +76,43 @@ public:
 
 private:
   llvm::SmallDenseMap<size_t, PimMemory> deviceMem;
+  std::fstream fileReport;
+  std::optional<MemoryReportRow> hostReportRow;
+  llvm::SmallVector<MemoryReportEntry, 32> reportEntries;
 
 public:
   PimAcceleratorMemory()
-  : hostMem(memEntriesMap) {}
+  : hostMem(memEntriesMap) {
+
+    std::string outputDir = getOutputDir();
+    if (outputDir.empty())
+      return;
+
+    std::string dialectsDir = outputDir + "/reports/";
+    createDirectory(dialectsDir);
+    std::fstream file(dialectsDir + "/memory_report.txt", std::ios::out);
+    fileReport = std::move(file);
+  }
 
   PimMemory& getOrCreateDeviceMem(size_t id);
 
   size_t getValueAddress(mlir::Value value, const StaticValueKnowledge& knowledge = {}) const;
+  void reportHost();
+  void recordCoreReport(size_t coreId, const MemoryReportRow& row);
+  void recordBatchReport(uint64_t batchId, llvm::ArrayRef<int32_t> coreIds, const MemoryReportRow& row);
+  void flushReport();
+  void clean(mlir::Operation* op);
 };
 
 class PimCodeGen {
   PimAcceleratorMemory& memory;
   llvm::raw_fd_ostream& coreFileStream;
+  const llvm::DenseMap<size_t, size_t>& emittedCoreIds;
 
   size_t addressOf(mlir::Value value, const StaticValueKnowledge& knowledge) const {
     return memory.getValueAddress(value, knowledge);
   }
+  size_t remapCoreId(size_t coreId) const;
 
   static llvm::json::Object createEmptyOffset();
   void emitInstruction(llvm::json::Object instruction) const;
@@ -83,15 +134,20 @@ class PimCodeGen {
   void emitMvmOp(size_t groupId, size_t rdAddr, size_t rdOffset, size_t rs1Addr, size_t rs1Offset) const;
 
 public:
-  PimCodeGen(PimAcceleratorMemory& memory, llvm::raw_fd_ostream& coreJson)
+  PimCodeGen(PimAcceleratorMemory& memory,
-  : memory(memory), coreFileStream(coreJson) {}
+             llvm::raw_fd_ostream& coreJson,
+             const llvm::DenseMap<size_t, size_t>& emittedCoreIds)
+  : memory(memory), coreFileStream(coreJson), emittedCoreIds(emittedCoreIds) {}
 
   void codeGenLoadOp(pim::PimMemCopyHostToDevOp loadOp, const StaticValueKnowledge& knowledge) const;
   void codeGenStoreOp(pim::PimMemCopyDevToHostOp storeOp, const StaticValueKnowledge& knowledge) const;
   void codeGenLmvOp(pim::PimMemCopyOp lmvOp, const StaticValueKnowledge& knowledge) const;
 
   void codeGenReceiveOp(pim::PimReceiveOp receiveOp, const StaticValueKnowledge& knowledge) const;
+  void codeGenReceiveTensorOp(pim::PimReceiveTensorOp receiveTensorOp, const StaticValueKnowledge& knowledge) const;
   void codeGenSendOp(pim::PimSendOp sendOp, const StaticValueKnowledge& knowledge) const;
+  void codeGenSendTensorOp(pim::PimSendTensorOp sendTensorOp, const StaticValueKnowledge& knowledge) const;
+  void codeGenConcatOp(pim::PimConcatOp concatOp, const StaticValueKnowledge& knowledge) const;
 
   template <typename MVMTy>
   void codeGenMVMLikeOp(size_t mvmId, MVMTy mvmLikeOp, bool transposeMatrix, const StaticValueKnowledge& knowledge);
@@ -106,6 +162,7 @@ public:
   void codeGenVTanhOp(pim::PimVTanhOp vtanhOp, const StaticValueKnowledge& knowledge) const;
   void codeGenVSigmOp(pim::PimVSigmOp vsigmOp, const StaticValueKnowledge& knowledge) const;
   void codeGenVSoftmaxOp(pim::PimVSoftmaxOp vsoftmaxOp, const StaticValueKnowledge& knowledge) const;
+  void codeGetGlobalOp(mlir::memref::GetGlobalOp getGlobalOp, const StaticValueKnowledge& knowledge) const;
   void codeGenTransposeOp(pim::PimTransposeOp transposeOp, const StaticValueKnowledge& knowledge) const;
 };
 

src/PIM/Compiler/PimCompilerOptions.cpp

@@ -1,16 +1,3 @@
-/*
- * SPDX-License-Identifier: Apache-2.0
- */
-
-//===------------------------- PimCompilerOptions.cpp --------------------===//
-//
-// Copyright 2022 The IBM Research Authors.
-//
-// =============================================================================
-//
-// Compiler Options for PIM
-//
-//===----------------------------------------------------------------------===//
 #include "src/Accelerators/PIM/Compiler/PimCompilerOptions.hpp"
 
 #define DEBUG_TYPE "PimCompilerOptions"
@@ -41,7 +28,7 @@ llvm::cl::opt<size_t>
   crossbarSize("crossbar-size", llvm::cl::desc("Width and heigth of a single crossbar"), llvm::cl::init(2));
 
 llvm::cl::opt<size_t>
-  crossbarCountInCore("crossbar-count", llvm::cl::desc("Number of crossbars in each core"), llvm::cl::init(2));
+  crossbarCountInCore("crossbar-count", llvm::cl::desc("Number of crossbars in each core"), llvm::cl::init(256));
 
 llvm::cl::opt<long> coresCount("core-count",
                                llvm::cl::desc("Number of cores in the chip. `-1` to use the minimum amount of cores."),
@@ -51,7 +38,7 @@ llvm::cl::opt<size_t> dcpCriticalWindowSize(
   "dcp-critical-window-size",
   llvm::cl::desc("Number of lowest-slack virtual nodes considered by each DCP coarsening iteration. "
                  "Use 0 to run the legacy full-graph DCP analysis."),
-  llvm::cl::init(1024));
+  llvm::cl::init(4000));
 
 llvm::cl::opt<bool>
   ignoreConcatError("ignore-concat-error",

src/PIM/Compiler/PimWeightEmitter.cpp

@@ -0,0 +1,220 @@
+#include "mlir/Dialect/MemRef/IR/MemRef.h"
+#include "mlir/IR/BuiltinAttributes.h"
+#include "mlir/IR/BuiltinOps.h"
+#include "mlir/IR/BuiltinTypes.h"
+
+#include "llvm/ADT/STLExtras.h"
+#include "llvm/Support/FileSystem.h"
+#include "llvm/Support/raw_ostream.h"
+
+#include <cassert>
+
+#include "Conversion/ONNXToSpatial/Common/Common.hpp"
+#include "src/Accelerators/PIM/Common/IR/WeightUtils.hpp"
+#include "src/Accelerators/PIM/Compiler/PimBatchEmission.hpp"
+#include "src/Accelerators/PIM/Compiler/PimCodeGen.hpp"
+#include "src/Accelerators/PIM/Compiler/PimCompilerOptions.hpp"
+#include "src/Accelerators/PIM/Compiler/PimWeightEmitter.hpp"
+#include "src/Accelerators/PIM/Dialect/Pim/PimOps.hpp"
+
+using namespace llvm;
+using namespace mlir;
+
+namespace onnx_mlir {
+namespace {
+
+struct DenseWeightView {
+  DenseElementsAttr denseAttr;
+  SmallVector<int64_t> shape;
+  SmallVector<int64_t> strides;
+  int64_t offset = 0;
+};
+
+SmallVector<int64_t> computeRowMajorStridesForShape(ArrayRef<int64_t> shape) {
+  SmallVector<int64_t> strides(shape.size(), 1);
+  for (int64_t index = static_cast<int64_t>(shape.size()) - 2; index >= 0; --index)
+    strides[index] = strides[index + 1] * shape[index + 1];
+  return strides;
+}
+
+bool allStaticSubviewParts(memref::SubViewOp subview) {
+  return llvm::all_of(subview.getStaticOffsets(), [](int64_t value) { return !ShapedType::isDynamic(value); })
+      && llvm::all_of(subview.getStaticSizes(), [](int64_t value) { return !ShapedType::isDynamic(value); })
+      && llvm::all_of(subview.getStaticStrides(), [](int64_t value) { return !ShapedType::isDynamic(value); });
+}
+
+FailureOr<DenseWeightView> resolveDenseWeightView(ModuleOp moduleOp, mlir::Value weight) {
+  SmallVector<memref::SubViewOp> subviews;
+  mlir::Value current = weight;
+  memref::GetGlobalOp getGlobalOp;
+
+  while (true) {
+    Operation* defOp = current.getDefiningOp();
+    if (!defOp)
+      return failure();
+    if ((getGlobalOp = dyn_cast<memref::GetGlobalOp>(defOp)))
+      break;
+    if (auto subview = dyn_cast<memref::SubViewOp>(defOp)) {
+      if (!allStaticSubviewParts(subview))
+        return failure();
+      subviews.push_back(subview);
+      current = subview.getSource();
+      continue;
+    }
+    if (auto cast = dyn_cast<memref::CastOp>(defOp)) {
+      current = cast.getSource();
+      continue;
+    }
+    return failure();
+  }
+
+  auto globalOp = lookupGlobalForGetGlobal(moduleOp, getGlobalOp);
+  if (!globalOp || !globalOp.getInitialValue())
+    return failure();
+
+  auto denseAttr = dyn_cast<DenseElementsAttr>(*globalOp.getInitialValue());
+  if (!denseAttr)
+    return failure();
+
+  DenseWeightView view;
+  view.denseAttr = denseAttr;
+  view.shape.assign(denseAttr.getType().getShape().begin(), denseAttr.getType().getShape().end());
+  view.strides = computeRowMajorStridesForShape(view.shape);
+
+  for (memref::SubViewOp subview : llvm::reverse(subviews)) {
+    SmallVector<int64_t> nextStrides;
+    nextStrides.reserve(subview.getStaticStrides().size());
+    for (auto [offset, stride, sourceStride] :
+         llvm::zip_equal(subview.getStaticOffsets(), subview.getStaticStrides(), view.strides)) {
+      view.offset += offset * sourceStride;
+      nextStrides.push_back(stride * sourceStride);
+    }
+    view.shape.assign(subview.getStaticSizes().begin(), subview.getStaticSizes().end());
+    view.strides = std::move(nextStrides);
+  }
+
+  return view;
+}
+
+SmallVector<unsigned, 8> getUsedWeightIndices(Block& block) {
+  SmallVector<unsigned, 8> indices;
+  auto addIndex = [&](unsigned weightIndex) {
+    if (!llvm::is_contained(indices, weightIndex))
+      indices.push_back(weightIndex);
+  };
+
+  block.walk([&](pim::PimVMMOp vmmOp) { addIndex(vmmOp.getWeightIndex()); });
+  llvm::sort(indices);
+  return indices;
+}
+
+SmallVector<unsigned, 8> getUsedWeightIndices(pim::PimCoreOp coreOp) {
+  return getUsedWeightIndices(coreOp.getBody().front());
+}
+
+SmallVector<Operation*> collectTopLevelCoreLikeOps(func::FuncOp funcOp) {
+  SmallVector<Operation*> coreLikeOps;
+  for (Operation& op : funcOp.getBody().front())
+    if (dyn_cast<pim::PimCoreOp>(&op) || dyn_cast<pim::PimCoreBatchOp>(&op))
+      coreLikeOps.push_back(&op);
+  return coreLikeOps;
+}
+
+} // namespace
+
+llvm::DenseMap<size_t, llvm::DenseMap<mlir::Value, std::string>>
+createAndPopulateWeightFolder(func::FuncOp funcOp, StringRef outputDirPath) {
+  ModuleOp moduleOp = funcOp->getParentOfType<ModuleOp>();
+  auto coreWeightsDirPath = outputDirPath + "/weights";
+  auto error = sys::fs::create_directory(coreWeightsDirPath);
+  assert(!error && "Error creating weights directory");
+  size_t indexFileName = 0;
+
+  int64_t xbarSize = crossbarSize.getValue();
+  llvm::DenseMap<size_t, llvm::DenseMap<mlir::Value, std::string>> mapCoreWeightToFileName;
+  llvm::DenseMap<memref::GlobalOp, std::string> mapGlobalOpToFileName;
+
+  SmallVector<Operation*> coreLikeOps = collectTopLevelCoreLikeOps(funcOp);
+
+  for (Operation* op : coreLikeOps) {
+    auto processCore = [&](pim::PimCoreOp coreOp) {
+      size_t coreId = static_cast<size_t>(coreOp.getCoreId());
+      for (unsigned index : getUsedWeightIndices(coreOp)) {
+        if (index >= coreOp.getWeights().size()) {
+          coreOp.emitWarning("Weight index " + std::to_string(index) + " is out of range");
+          assert(index < coreOp.getWeights().size() && "Weight index is out of range");
+        }
+        mlir::Value weight = coreOp.getWeights()[index];
+
+        auto weightView = resolveDenseWeightView(moduleOp, weight);
+        if (failed(weightView)) {
+          coreOp.emitWarning("Weight is not from a memref.get_global at index " + std::to_string(index));
+          assert(succeeded(weightView) && "Weight is not from a dense memref.global view");
+        }
+
+        if (mapCoreWeightToFileName[coreId].contains(weight))
+          continue;
+
+        auto getGlobalOp = weight.getDefiningOp<memref::GetGlobalOp>();
+        auto globalOp = getGlobalOp ? lookupGlobalForGetGlobal(moduleOp, getGlobalOp) : memref::GlobalOp {};
+        if (globalOp && mapGlobalOpToFileName.contains(globalOp)) {
+          auto& fileName = mapGlobalOpToFileName[globalOp];
+          mapCoreWeightToFileName[coreId].insert({weight, fileName});
+          continue;
+        }
+
+        DenseElementsAttr denseAttr = weightView->denseAttr;
+        ArrayRef<int64_t> shape = weightView->shape;
+        assert(isMatrixShape(shape) && "Weight matrix must be 2-dimensional");
+        int64_t numRows = shape[0];
+        int64_t numCols = shape[1];
+        assert(numRows <= xbarSize && numCols <= xbarSize && "Weight dimensions must not exceed crossbar size");
+
+        size_t elementByteWidth = denseAttr.getElementType().getIntOrFloatBitWidth() / 8;
+
+        std::string newFileName = "crossbar_" + std::to_string(indexFileName++) + ".bin";
+        auto weightFilePath = (coreWeightsDirPath + "/" + newFileName).str();
+        std::error_code errorCode;
+        raw_fd_ostream weightFileStream(weightFilePath, errorCode, sys::fs::OF_None);
+        if (errorCode) {
+          errs() << "Error while opening weight file `" << weightFilePath << "`: " << errorCode.message() << '\n';
+          assert(errorCode);
+        }
+
+        uint64_t zero = 0;
+        for (int64_t row = 0; row < xbarSize; row++) {
+          for (int64_t col = 0; col < xbarSize; col++) {
+            if (row < numRows && col < numCols) {
+              int64_t elementIndex = weightView->offset + row * weightView->strides[0] + col * weightView->strides[1];
+              APInt bits = denseAttr.getValues<APFloat>()[elementIndex].bitcastToAPInt();
+              uint64_t word = bits.getZExtValue();
+              weightFileStream.write(reinterpret_cast<const char*>(&word), elementByteWidth);
+            }
+            else {
+              weightFileStream.write(reinterpret_cast<const char*>(&zero), elementByteWidth);
+            }
+          }
+        }
+
+        weightFileStream.close();
+        if (globalOp)
+          mapGlobalOpToFileName.insert({globalOp, newFileName});
+        mapCoreWeightToFileName[coreId].insert({weight, newFileName});
+      }
+      return success();
+    };
+
+    if (auto coreOp = dyn_cast<pim::PimCoreOp>(op)) {
+      (void) processCore(coreOp);
+      continue;
+    }
+
+    auto coreBatchOp = cast<pim::PimCoreBatchOp>(op);
+    for (unsigned lane = 0; lane < static_cast<unsigned>(coreBatchOp.getLaneCount()); ++lane)
+      if (failed(withScalarCoreFromBatchLane(coreBatchOp, lane, processCore)))
+        return mapCoreWeightToFileName;
+  }
+  return mapCoreWeightToFileName;
+}
+
+} // namespace onnx_mlir

src/PIM/Compiler/PimWeightEmitter.hpp

@@ -0,0 +1,16 @@
+#pragma once
+
+#include "mlir/Dialect/Func/IR/FuncOps.h"
+#include "mlir/IR/Value.h"
+
+#include "llvm/ADT/DenseMap.h"
+#include "llvm/ADT/StringRef.h"
+
+#include <string>
+
+namespace onnx_mlir {
+
+llvm::DenseMap<size_t, llvm::DenseMap<mlir::Value, std::string>>
+createAndPopulateWeightFolder(mlir::func::FuncOp funcOp, llvm::StringRef outputDirPath);
+
+} // namespace onnx_mlir

src/PIM/Conversion/ONNXToSpatial/CMakeLists.txt

@@ -3,6 +3,11 @@ mlir_tablegen(ONNXToSpatial.hpp.inc -gen-rewriters "-I${ONNX_MLIR_SRC_ROOT}")
 add_public_tablegen_target(ONNXToSpatialIncGen)
 
 add_pim_library(OMONNXToSpatial
+  ConversionPatterns.cpp
+  HostFoldability.cpp
+  HostLegality.cpp
+  PrePatterns.cpp
+  PostPatterns.cpp
   Patterns/Math/Conv.cpp
   Patterns/Math/Elementwise.cpp
   Patterns/Math/Gemm.cpp
@@ -18,7 +23,9 @@ add_pim_library(OMONNXToSpatial
   Patterns/Tensor/Reshape.cpp
   Patterns/Tensor/Split.cpp
   ONNXToSpatialPass.cpp
-  Common.cpp
+  Common/ComputeRegionBuilder.cpp
+  Common/ShapeTilingUtils.cpp
+  Common/WeightMaterialization.cpp
 
   EXCLUDE_FROM_OM_LIBS
 

src/PIM/Conversion/ONNXToSpatial/Common.hpp

@@ -1,279 +0,0 @@
-#pragma once
-
-#include "mlir/Dialect/Arith/IR/Arith.h"
-#include "mlir/Dialect/Tensor/IR/Tensor.h"
-#include "mlir/IR/Block.h"
-#include "mlir/IR/BuiltinTypes.h"
-#include "mlir/IR/ValueRange.h"
-#include "mlir/Transforms/DialectConversion.h"
-
-#include <cassert>
-#include <type_traits>
-#include <utility>
-
-#include "llvm/ADT/SmallPtrSet.h"
-
-#include "src/Accelerators/PIM/Common/PimCommon.hpp"
-#include "src/Accelerators/PIM/Dialect/Spatial/SpatialOps.hpp"
-#include "src/Dialect/ONNX/ONNXOps.hpp"
-
-namespace onnx_mlir {
-
-template <class ShapedType>
-inline auto getImageWidth(const ShapedType& shapedType) {
-  return shapedType.getDimSize(2);
-}
-
-template <class ShapedType>
-inline auto getImageHeight(const ShapedType& shapedType) {
-  return shapedType.getDimSize(3);
-}
-
-template <class ShapedType>
-inline auto getImageChannel(const ShapedType& shapedType) {
-  return shapedType.getDimSize(1);
-}
-
-template <class ShapedType>
-inline auto getImageN(const ShapedType& shapedType) {
-  return shapedType.getDimSize(0);
-}
-
-template <class ShapedType>
-inline auto getKernelWidth(const ShapedType& shapedType) {
-  return shapedType.getDimSize(2);
-}
-
-template <class ShapedType>
-inline auto getKernelHeight(const ShapedType& shapedType) {
-  return shapedType.getDimSize(3);
-}
-
-template <class ShapedType>
-inline auto getFilterCount(const ShapedType& shapedType) {
-  return shapedType.getDimSize(0);
-}
-
-using HSliceId = size_t;
-using CoreId = size_t;
-
-template <class A, class B, class C = std::common_type_t<A, B>>
-constexpr C ceilIntegerDivide(A a, B b) {
-  static_assert(std::is_integral_v<A>, "A must be an integer type");
-  static_assert(std::is_integral_v<B>, "B must be an integer type");
-  C ac = static_cast<C>(a);
-  C bc = static_cast<C>(b);
-  return 1 + (ac - 1) / bc;
-}
-
-template <class A, class B, class C = std::common_type_t<A, B>>
-constexpr std::pair<C, C> ceilIntegerDivideWithRemainder(A a, B b) {
-  static_assert(std::is_integral_v<A>, "A must be an integer type");
-  static_assert(std::is_integral_v<B>, "B must be an integer type");
-  C ac = static_cast<C>(a);
-  C bc = static_cast<C>(b);
-  return {ceilIntegerDivide(ac, bc), ac % bc};
-}
-
-template <class T>
-bool isVectorShape(mlir::ArrayRef<T> shape) {
-  return shape.size() == 2 && (shape[0] == 1 || shape[1] == 1);
-}
-
-template <class T>
-bool isMatrixShape(mlir::ArrayRef<T> shape) {
-  return shape.size() == 2;
-}
-
-template <class T>
-bool isHVectorShape(mlir::ArrayRef<T> shape) {
-  return shape.size() == 2 && shape[0] == 1;
-}
-
-template <class T>
-bool isVVectorShape(mlir::ArrayRef<T> shape) {
-  return shape.size() == 2 && shape[1] == 1;
-}
-
-template <class T>
-T getVectorLength(mlir::ArrayRef<T> shape) {
-  assert(isVectorShape(shape));
-  return shape[0] != 1 ? shape[0] : shape[1];
-}
-
-inline auto getTensorShape(mlir::Value tensor) {
-  return mlir::cast<mlir::RankedTensorType>(tensor.getType()).getShape();
-}
-
-inline bool isWeightLikeComputeOperand(mlir::Value value) {
-  auto rankedType = mlir::dyn_cast<mlir::RankedTensorType>(value.getType());
-  if (!rankedType || !isMatrixShape(rankedType.getShape()))
-    return false;
-
-  llvm::SmallPtrSet<mlir::Operation*, 8> visited;
-
-  while (auto* definingOp = value.getDefiningOp()) {
-    if (!visited.insert(definingOp).second)
-      return false;
-    if (hasWeightAlways(definingOp))
-      return true;
-
-    if (auto extractSliceOp = mlir::dyn_cast<mlir::tensor::ExtractSliceOp>(definingOp)) {
-      value = extractSliceOp.getSource();
-      continue;
-    }
-    if (auto expandShapeOp = mlir::dyn_cast<mlir::tensor::ExpandShapeOp>(definingOp)) {
-      value = expandShapeOp.getSrc();
-      continue;
-    }
-    if (auto collapseShapeOp = mlir::dyn_cast<mlir::tensor::CollapseShapeOp>(definingOp)) {
-      value = collapseShapeOp.getSrc();
-      continue;
-    }
-    if (auto transposeOp = mlir::dyn_cast<mlir::ONNXTransposeOp>(definingOp)) {
-      value = transposeOp.getData();
-      continue;
-    }
-
-    return false;
-  }
-
-  return false;
-}
-
-namespace detail {
-
-inline mlir::ValueRange getBlockArgs(mlir::Block* block) { return mlir::ValueRange(block->getArguments()); }
-
-template <typename Fn, size_t... Is>
-decltype(auto) invokeWithBlockArgs(Fn&& fn, mlir::Block* block, std::index_sequence<Is...>) {
-  return std::forward<Fn>(fn)(block->getArgument(Is)...);
-}
-
-template <typename Fn, size_t... Is>
-decltype(auto) invokeWithValues(Fn&& fn, mlir::ArrayRef<mlir::Value> values, std::index_sequence<Is...>) {
-  return std::forward<Fn>(fn)(values[Is]...);
-}
-
-template <size_t>
-using ValueArg = mlir::Value;
-
-template <typename Fn, typename Seq>
-struct InvokeWithBlockArgsResult;
-
-template <typename Fn, size_t... Is>
-struct InvokeWithBlockArgsResult<Fn, std::index_sequence<Is...>> {
-  using type = std::invoke_result_t<Fn, ValueArg<Is>...>;
-};
-
-template <typename Fn, typename Seq>
-using InvokeWithBlockArgsResultT = typename InvokeWithBlockArgsResult<Fn, Seq>::type;
-
-template <typename Fn>
-using InvokeWithValueRangeResultT = std::invoke_result_t<Fn, mlir::ValueRange>;
-
-} // namespace detail
-
-template <size_t NumInputs, typename RewriterT, typename BodyFn>
-auto createSpatCompute(RewriterT& rewriter,
-                       mlir::Location loc,
-                       mlir::TypeRange resultTypes,
-                       mlir::ValueRange weights,
-                       mlir::ValueRange inputs,
-                       BodyFn&& body) {
-  assert(inputs.size() == NumInputs && "NumInputs must match the number of input values");
-  auto computeOp = spatial::SpatCompute::create(rewriter, loc, resultTypes, weights, inputs);
-
-  auto* block = new mlir::Block();
-  for (mlir::Value input : inputs)
-    block->addArgument(input.getType(), loc);
-
-  computeOp.getBody().push_back(block);
-  rewriter.setInsertionPointToStart(block);
-
-  using BodyResult = detail::InvokeWithBlockArgsResultT<std::decay_t<BodyFn>, std::make_index_sequence<NumInputs>>;
-  if constexpr (std::is_same_v<BodyResult, mlir::LogicalResult>) {
-    auto bodyResult =
-      detail::invokeWithBlockArgs(std::forward<BodyFn>(body), block, std::make_index_sequence<NumInputs> {});
-    if (mlir::failed(bodyResult)) {
-      rewriter.setInsertionPointAfter(computeOp);
-      rewriter.eraseOp(computeOp);
-      return mlir::FailureOr<spatial::SpatCompute>(mlir::failure());
-    }
-    rewriter.setInsertionPointAfter(computeOp);
-    return mlir::FailureOr<spatial::SpatCompute>(computeOp);
-  }
-  else {
-    static_assert(std::is_same_v<BodyResult, void>, "createSpatCompute body must return void or mlir::LogicalResult");
-    detail::invokeWithBlockArgs(std::forward<BodyFn>(body), block, std::make_index_sequence<NumInputs> {});
-
-    rewriter.setInsertionPointAfter(computeOp);
-    return computeOp;
-  }
-}
-
-template <typename RewriterT, typename BodyFn>
-auto createSpatCompute(RewriterT& rewriter,
-                       mlir::Location loc,
-                       mlir::TypeRange resultTypes,
-                       mlir::ValueRange weights,
-                       mlir::ValueRange inputs,
-                       BodyFn&& body) {
-  auto computeOp = spatial::SpatCompute::create(rewriter, loc, resultTypes, weights, inputs);
-
-  auto* block = new mlir::Block();
-  for (mlir::Value input : inputs)
-    block->addArgument(input.getType(), loc);
-
-  computeOp.getBody().push_back(block);
-  rewriter.setInsertionPointToStart(block);
-
-  using BodyResult = detail::InvokeWithValueRangeResultT<std::decay_t<BodyFn>>;
-  if constexpr (std::is_same_v<BodyResult, mlir::LogicalResult>) {
-    auto bodyResult = std::forward<BodyFn>(body)(detail::getBlockArgs(block));
-    if (mlir::failed(bodyResult)) {
-      rewriter.setInsertionPointAfter(computeOp);
-      rewriter.eraseOp(computeOp);
-      return mlir::FailureOr<spatial::SpatCompute>(mlir::failure());
-    }
-    rewriter.setInsertionPointAfter(computeOp);
-    return mlir::FailureOr<spatial::SpatCompute>(computeOp);
-  }
-  else {
-    static_assert(std::is_same_v<BodyResult, void>, "createSpatCompute body must return void or mlir::LogicalResult");
-    std::forward<BodyFn>(body)(detail::getBlockArgs(block));
-
-    rewriter.setInsertionPointAfter(computeOp);
-    return computeOp;
-  }
-}
-
-llvm::SmallVector<mlir::Value> sliceTensor(const mlir::Value& tensorToSlice,
-                                           size_t axis,
-                                           int64_t sliceSize,
-                                           mlir::ConversionPatternRewriter& rewriter,
-                                           mlir::Location loc);
-
-llvm::SmallVector<mlir::Value> sliceVector(const mlir::Value& vectorToSlice,
-                                           int64_t sliceSize,
-                                           mlir::ConversionPatternRewriter& rewriter,
-                                           mlir::Location loc);
-
-llvm::DenseMap<CoreId, llvm::SmallVector<mlir::Value>> sliceVectorPerCrossbarPerCore(
-  const mlir::Value& vectorToSlice, mlir::ConversionPatternRewriter& rewriter, mlir::Location loc);
-
-llvm::DenseMap<HSliceId, llvm::DenseMap<CoreId, llvm::SmallVector<mlir::Value>>>
-tileMatrix(mlir::Value& matrixToTile,
-           int64_t hSliceSize,
-           int64_t vSliceSize,
-           mlir::ConversionPatternRewriter& rewriter,
-           mlir::Location& loc);
-
-mlir::tensor::SplatOp broadcastToVector(mlir::Value scalarToBroadcast,
-                                        int64_t length,
-                                        mlir::ConversionPatternRewriter& rewriter,
-                                        mlir::Location loc);
-
-mlir::Value sumTensors(mlir::ArrayRef<mlir::Value> tensors, mlir::ConversionPatternRewriter& rewriter);
-
-}; // namespace onnx_mlir

src/PIM/Conversion/ONNXToSpatial/Common/Common.hpp

@@ -0,0 +1,7 @@
+#pragma once
+
+#include "ComputeRegionBuilder.hpp"
+#include "ShapeTilingUtils.hpp"
+#include "WeightMaterialization.hpp"
+#include "src/Accelerators/PIM/Common/PimCommon.hpp"
+#include "src/Dialect/ONNX/ONNXOps.hpp"

src/PIM/Conversion/ONNXToSpatial/Common/ComputeRegionBuilder.cpp

@@ -0,0 +1,39 @@
+#include "mlir/Transforms/DialectConversion.h"
+
+#include "llvm/ADT/SmallVector.h"
+
+#include "ComputeRegionBuilder.hpp"
+#include "src/Accelerators/PIM/Dialect/Spatial/SpatialOps.hpp"
+
+using namespace mlir;
+
+namespace onnx_mlir {
+
+Value sumTensors(ArrayRef<Value> tensors, ConversionPatternRewriter& rewriter) {
+  if (tensors.size() == 1)
+    return tensors[0];
+
+  SmallVector<Value> tensors1 = {tensors.begin(), tensors.end()};
+  SmallVector<Value> tensors2;
+  tensors2.reserve(tensors.size() / 2);
+
+  auto* currTensors = &tensors1;
+  auto* nextTensors = &tensors2;
+  while (currTensors->size() > 1) {
+    for (size_t i = 0; i < currTensors->size() - 1; i += 2) {
+      Value a = (*currTensors)[i];
+      Value b = (*currTensors)[i + 1];
+      rewriter.setInsertionPointAfterValue(b);
+      auto addedValue = spatial::SpatVAddOp::create(rewriter, a.getLoc(), a.getType(), a, b);
+      nextTensors->push_back(addedValue);
+    }
+    if (currTensors->size() % 2 == 1)
+      nextTensors->push_back(currTensors->back());
+    std::swap(currTensors, nextTensors);
+    nextTensors->clear();
+  }
+  assert(currTensors->size() == 1 && "Expected a single input at this point.");
+  return (*currTensors)[0];
+}
+
+} // namespace onnx_mlir

src/PIM/Conversion/ONNXToSpatial/Common/ComputeRegionBuilder.hpp

@@ -0,0 +1,153 @@
+#pragma once
+
+#include "mlir/IR/Block.h"
+#include "mlir/IR/BuiltinTypes.h"
+#include "mlir/IR/ValueRange.h"
+#include "mlir/Transforms/DialectConversion.h"
+
+#include <cassert>
+#include <cstddef>
+#include <type_traits>
+#include <utility>
+
+#include "src/Accelerators/PIM/Dialect/Spatial/SpatialOps.hpp"
+
+namespace onnx_mlir {
+
+namespace detail {
+
+inline mlir::ValueRange getBlockArgs(mlir::Block* block) { return mlir::ValueRange(block->getArguments()); }
+
+template <typename Fn, size_t... Is>
+decltype(auto) invokeWithBlockArgs(Fn&& fn, mlir::Block* block, std::index_sequence<Is...>) {
+  return std::forward<Fn>(fn)(block->getArgument(Is)...);
+}
+
+template <typename Fn, size_t... Is>
+decltype(auto) invokeWithValues(Fn&& fn, mlir::ArrayRef<mlir::Value> values, std::index_sequence<Is...>) {
+  return std::forward<Fn>(fn)(values[Is]...);
+}
+
+template <size_t>
+using ValueArg = mlir::Value;
+
+template <typename Fn, typename Seq>
+struct InvokeWithBlockArgsResult;
+
+template <typename Fn, size_t... Is>
+struct InvokeWithBlockArgsResult<Fn, std::index_sequence<Is...>> {
+  using type = std::invoke_result_t<Fn, ValueArg<Is>...>;
+};
+
+template <typename Fn, typename Seq>
+using InvokeWithBlockArgsResultT = typename InvokeWithBlockArgsResult<Fn, Seq>::type;
+
+template <typename Fn>
+using InvokeWithValueRangeResultT = std::invoke_result_t<Fn, mlir::ValueRange>;
+
+} // namespace detail
+
+template <typename RewriterT>
+inline mlir::Value createSpatConcat(RewriterT& rewriter, mlir::Location loc, int64_t axis, mlir::ValueRange inputs) {
+  assert(!inputs.empty() && "spat.concat requires at least one input");
+  if (inputs.size() == 1)
+    return inputs.front();
+
+  auto firstType = mlir::cast<mlir::RankedTensorType>(inputs.front().getType());
+  auto outputShape = llvm::to_vector(firstType.getShape());
+  int64_t concatDimSize = 0;
+  bool concatDimDynamic = false;
+
+  for (mlir::Value input : inputs) {
+    auto inputType = mlir::cast<mlir::RankedTensorType>(input.getType());
+    assert(inputType.getRank() == firstType.getRank() && "spat.concat expects same-rank inputs");
+    if (mlir::ShapedType::isDynamic(inputType.getDimSize(axis)))
+      concatDimDynamic = true;
+    else
+      concatDimSize += inputType.getDimSize(axis);
+  }
+
+  outputShape[axis] = concatDimDynamic ? mlir::ShapedType::kDynamic : concatDimSize;
+  auto outputType = mlir::RankedTensorType::get(outputShape, firstType.getElementType(), firstType.getEncoding());
+  return spatial::SpatConcatOp::create(rewriter, loc, outputType, rewriter.getI64IntegerAttr(axis), inputs).getOutput();
+}
+
+/// Builds a `spat.compute` with a fixed number of SSA inputs and erases it if
+/// the body callback reports failure.
+template <size_t NumInputs, typename RewriterT, typename BodyFn>
+auto createSpatCompute(RewriterT& rewriter,
+                       mlir::Location loc,
+                       mlir::TypeRange resultTypes,
+                       mlir::ValueRange weights,
+                       mlir::ValueRange inputs,
+                       BodyFn&& body) {
+  assert(inputs.size() == NumInputs && "NumInputs must match the number of input values");
+  auto computeOp = spatial::SpatCompute::create(rewriter, loc, resultTypes, weights, inputs);
+
+  auto* block = new mlir::Block();
+  for (mlir::Value input : inputs)
+    block->addArgument(input.getType(), loc);
+
+  computeOp.getBody().push_back(block);
+  rewriter.setInsertionPointToStart(block);
+
+  using BodyResult = detail::InvokeWithBlockArgsResultT<std::decay_t<BodyFn>, std::make_index_sequence<NumInputs>>;
+  if constexpr (std::is_same_v<BodyResult, void>) {
+    detail::invokeWithBlockArgs(std::forward<BodyFn>(body), block, std::make_index_sequence<NumInputs> {});
+
+    rewriter.setInsertionPointAfter(computeOp);
+    return computeOp;
+  }
+  else {
+    auto bodyResult =
+      detail::invokeWithBlockArgs(std::forward<BodyFn>(body), block, std::make_index_sequence<NumInputs> {});
+    if (mlir::failed(bodyResult)) {
+      rewriter.setInsertionPointAfter(computeOp);
+      rewriter.eraseOp(computeOp);
+      return mlir::FailureOr<spatial::SpatCompute>(mlir::failure());
+    }
+    rewriter.setInsertionPointAfter(computeOp);
+    return mlir::FailureOr<spatial::SpatCompute>(computeOp);
+  }
+}
+
+/// Builds a `spat.compute` whose body consumes the block arguments as a single
+/// `ValueRange`, which is convenient for variadic reductions/concats.
+template <typename RewriterT, typename BodyFn>
+auto createSpatCompute(RewriterT& rewriter,
+                       mlir::Location loc,
+                       mlir::TypeRange resultTypes,
+                       mlir::ValueRange weights,
+                       mlir::ValueRange inputs,
+                       BodyFn&& body) {
+  auto computeOp = spatial::SpatCompute::create(rewriter, loc, resultTypes, weights, inputs);
+
+  auto* block = new mlir::Block();
+  for (mlir::Value input : inputs)
+    block->addArgument(input.getType(), loc);
+
+  computeOp.getBody().push_back(block);
+  rewriter.setInsertionPointToStart(block);
+
+  using BodyResult = detail::InvokeWithValueRangeResultT<std::decay_t<BodyFn>>;
+  if constexpr (std::is_same_v<BodyResult, void>) {
+    std::forward<BodyFn>(body)(detail::getBlockArgs(block));
+
+    rewriter.setInsertionPointAfter(computeOp);
+    return computeOp;
+  }
+  else {
+    auto bodyResult = std::forward<BodyFn>(body)(detail::getBlockArgs(block));
+    if (mlir::failed(bodyResult)) {
+      rewriter.setInsertionPointAfter(computeOp);
+      rewriter.eraseOp(computeOp);
+      return mlir::FailureOr<spatial::SpatCompute>(mlir::failure());
+    }
+    rewriter.setInsertionPointAfter(computeOp);
+    return mlir::FailureOr<spatial::SpatCompute>(computeOp);
+  }
+}
+
+mlir::Value sumTensors(mlir::ArrayRef<mlir::Value> tensors, mlir::ConversionPatternRewriter& rewriter);
+
+} // namespace onnx_mlir

src/PIM/Conversion/ONNXToSpatial/Common/ShapeTilingUtils.cpp

@@ -1,24 +1,12 @@
+#include "mlir/Dialect/Arith/IR/Arith.h"
 #include "mlir/Dialect/Tensor/IR/Tensor.h"
-#include "mlir/Dialect/Tosa/IR/TosaOps.h"
-#include "mlir/IR/BuiltinAttributes.h"
-#include "mlir/IR/BuiltinTypes.h"
-#include "mlir/IR/Location.h"
-#include "mlir/IR/PatternMatch.h"
-#include "mlir/IR/Value.h"
-#include "mlir/Transforms/DialectConversion.h"
 
 #include "llvm/ADT/SmallVector.h"
-#include "llvm/ADT/Twine.h"
-#include "llvm/Support/Casting.h"
 
-#include <cassert>
+#include "ShapeTilingUtils.hpp"
-#include <optional>
-#include <utility>
-
-#include "Common.hpp"
 #include "src/Accelerators/PIM/Compiler/PimCompilerOptions.hpp"
-#include "src/Accelerators/PIM/Dialect/Spatial/SpatialOps.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Common/Common.hpp"
-#include "src/Dialect/ONNX/ONNXOps.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/HostFoldability.hpp"
 
 using namespace mlir;
 
@@ -44,10 +32,29 @@ SmallVector<Value> sliceTensor(
 
   for (int64_t i = 0; i < numSlices; i++) {
     offsets[axis] = rewriter.getIndexAttr(i * sliceSize);
-    if (i == numSlices - 1 && lastSliceSize != 0)
+    int64_t currentSliceSize = sliceSize;
+    if (i == numSlices - 1 && lastSliceSize != 0) {
+      currentSliceSize = lastSliceSize;
       sizes[axis] = rewriter.getIndexAttr(lastSliceSize);
+    }
 
-    Value slice = tensor::ExtractSliceOp::create(rewriter, loc, tensorToSlice, offsets, sizes, strides);
+    SmallVector<int64_t> sliceShape(shape.begin(), shape.end());
+    sliceShape[axis] = currentSliceSize;
+    auto sliceType =
+      RankedTensorType::get(sliceShape, cast<RankedTensorType>(tensorToSlice.getType()).getElementType());
+
+    Value slice;
+    if (isHostFoldableValue(tensorToSlice)) {
+      slice = tensor::ExtractSliceOp::create(rewriter, loc, tensorToSlice, offsets, sizes, strides);
+    }
+    else {
+      auto sliceCompute =
+        createSpatCompute<1>(rewriter, loc, TypeRange {sliceType}, {}, ValueRange {tensorToSlice}, [&](Value input) {
+          Value computedSlice = tensor::ExtractSliceOp::create(rewriter, loc, input, offsets, sizes, strides);
+          spatial::SpatYieldOp::create(rewriter, loc, computedSlice);
+        });
+      slice = sliceCompute.getResult(0);
+    }
     slices.push_back(slice);
   }
 
@@ -107,31 +114,4 @@ broadcastToVector(Value scalarToBroadcast, int64_t length, ConversionPatternRewr
   return tensor::SplatOp::create(rewriter, loc, type, elementValue);
 }
 
-Value sumTensors(ArrayRef<Value> tensors, ConversionPatternRewriter& rewriter) {
+} // namespace onnx_mlir
-  if (tensors.size() == 1)
-    return tensors[0];
-
-  SmallVector<Value> tensors1 = {tensors.begin(), tensors.end()};
-  SmallVector<Value> tensors2;
-  tensors2.reserve(tensors.size() / 2);
-
-  auto* currTensors = &tensors1;
-  auto* nextTensors = &tensors2;
-  while (currTensors->size() > 1) {
-    for (size_t i = 0; i < currTensors->size() - 1; i += 2) {
-      Value a = (*currTensors)[i];
-      Value b = (*currTensors)[i + 1];
-      rewriter.setInsertionPointAfterValue(b);
-      auto addedValue = spatial::SpatVAddOp::create(rewriter, a.getLoc(), a.getType(), a, b);
-      nextTensors->push_back(addedValue);
-    }
-    if (currTensors->size() % 2 == 1)
-      nextTensors->push_back(currTensors->back());
-    std::swap(currTensors, nextTensors);
-    nextTensors->clear();
-  }
-  assert(currTensors->size() == 1 && "Expected a single input at this point.");
-  return (*currTensors)[0];
-}
-
-}; // namespace onnx_mlir

src/PIM/Conversion/ONNXToSpatial/Common/ShapeTilingUtils.hpp

@@ -0,0 +1,144 @@
+#pragma once
+
+#include "mlir/Dialect/Tensor/IR/Tensor.h"
+#include "mlir/IR/BuiltinTypes.h"
+#include "mlir/IR/Value.h"
+#include "mlir/Transforms/DialectConversion.h"
+
+#include "llvm/ADT/ArrayRef.h"
+#include "llvm/ADT/DenseMap.h"
+#include "llvm/ADT/SmallVector.h"
+
+#include <cassert>
+#include <cstddef>
+#include <type_traits>
+#include <utility>
+
+namespace onnx_mlir {
+
+template <class ShapedType>
+inline auto getImageWidth(const ShapedType& shapedType) {
+  return shapedType.getDimSize(2);
+}
+
+template <class ShapedType>
+inline auto getImageHeight(const ShapedType& shapedType) {
+  return shapedType.getDimSize(3);
+}
+
+template <class ShapedType>
+inline auto getImageChannel(const ShapedType& shapedType) {
+  return shapedType.getDimSize(1);
+}
+
+template <class ShapedType>
+inline auto getImageN(const ShapedType& shapedType) {
+  return shapedType.getDimSize(0);
+}
+
+template <class ShapedType>
+inline auto getKernelWidth(const ShapedType& shapedType) {
+  return shapedType.getDimSize(2);
+}
+
+template <class ShapedType>
+inline auto getKernelHeight(const ShapedType& shapedType) {
+  return shapedType.getDimSize(3);
+}
+
+template <class ShapedType>
+inline auto getFilterCount(const ShapedType& shapedType) {
+  return shapedType.getDimSize(0);
+}
+
+using HSliceId = size_t;
+using CoreId = size_t;
+
+template <class A, class B, class C = std::common_type_t<A, B>>
+constexpr C ceilIntegerDivide(A a, B b) {
+  static_assert(std::is_integral_v<A>, "A must be an integer type");
+  static_assert(std::is_integral_v<B>, "B must be an integer type");
+  C ac = static_cast<C>(a);
+  C bc = static_cast<C>(b);
+  return 1 + (ac - 1) / bc;
+}
+
+template <class A, class B, class C = std::common_type_t<A, B>>
+constexpr std::pair<C, C> ceilIntegerDivideWithRemainder(A a, B b) {
+  static_assert(std::is_integral_v<A>, "A must be an integer type");
+  static_assert(std::is_integral_v<B>, "B must be an integer type");
+  C ac = static_cast<C>(a);
+  C bc = static_cast<C>(b);
+  return {ceilIntegerDivide(ac, bc), ac % bc};
+}
+
+template <class T>
+bool isVectorShape(mlir::ArrayRef<T> shape) {
+  return shape.size() == 2 && (shape[0] == 1 || shape[1] == 1);
+}
+
+template <class T>
+bool isMatrixShape(mlir::ArrayRef<T> shape) {
+  return shape.size() == 2;
+}
+
+template <class T>
+bool isHVectorShape(mlir::ArrayRef<T> shape) {
+  return shape.size() == 2 && shape[0] == 1;
+}
+
+template <class T>
+bool isVVectorShape(mlir::ArrayRef<T> shape) {
+  return shape.size() == 2 && shape[1] == 1;
+}
+
+template <class T>
+T getVectorLength(mlir::ArrayRef<T> shape) {
+  assert(isVectorShape(shape));
+  return shape[0] != 1 ? shape[0] : shape[1];
+}
+
+inline auto getTensorShape(mlir::Value tensor) {
+  return mlir::cast<mlir::RankedTensorType>(tensor.getType()).getShape();
+}
+
+inline bool haveSameStaticShape(mlir::Value lhs, mlir::Value rhs) {
+  auto lhsType = mlir::dyn_cast<mlir::RankedTensorType>(lhs.getType());
+  auto rhsType = mlir::dyn_cast<mlir::RankedTensorType>(rhs.getType());
+  return lhsType && rhsType && lhsType.hasStaticShape() && rhsType.hasStaticShape()
+      && lhsType.getShape() == rhsType.getShape();
+}
+
+/// Slices a statically shaped tensor along one axis into contiguous pieces of
+/// at most `sliceSize` elements.
+llvm::SmallVector<mlir::Value> sliceTensor(const mlir::Value& tensorToSlice,
+                                           size_t axis,
+                                           int64_t sliceSize,
+                                           mlir::ConversionPatternRewriter& rewriter,
+                                           mlir::Location loc);
+
+llvm::SmallVector<mlir::Value> sliceVector(const mlir::Value& vectorToSlice,
+                                           int64_t sliceSize,
+                                           mlir::ConversionPatternRewriter& rewriter,
+                                           mlir::Location loc);
+
+/// Partitions one logical vector into per-core crossbar-sized slices using the
+/// current PIM target geometry.
+llvm::DenseMap<CoreId, llvm::SmallVector<mlir::Value>> sliceVectorPerCrossbarPerCore(
+  const mlir::Value& vectorToSlice, mlir::ConversionPatternRewriter& rewriter, mlir::Location loc);
+
+/// Tiles a matrix first across output columns and then across input rows so it
+/// can be assigned to crossbars grouped by core.
+llvm::DenseMap<HSliceId, llvm::DenseMap<CoreId, llvm::SmallVector<mlir::Value>>>
+tileMatrix(mlir::Value& matrixToTile,
+           int64_t hSliceSize,
+           int64_t vSliceSize,
+           mlir::ConversionPatternRewriter& rewriter,
+           mlir::Location& loc);
+
+mlir::tensor::SplatOp broadcastToVector(mlir::Value scalarToBroadcast,
+                                        int64_t length,
+                                        mlir::ConversionPatternRewriter& rewriter,
+                                        mlir::Location loc);
+
+} // namespace onnx_mlir

src/PIM/Conversion/ONNXToSpatial/Common/WeightMaterialization.cpp

@@ -0,0 +1,114 @@
+#include "mlir/Dialect/Arith/IR/Arith.h"
+#include "mlir/Dialect/Tensor/IR/Tensor.h"
+#include "mlir/IR/BuiltinTypes.h"
+#include "mlir/IR/IRMapping.h"
+#include "mlir/IR/Value.h"
+#include "mlir/Support/LogicalResult.h"
+
+#include "llvm/ADT/STLExtras.h"
+#include "llvm/ADT/SmallPtrSet.h"
+#include "llvm/ADT/SmallVector.h"
+
+#include "ShapeTilingUtils.hpp"
+#include "WeightMaterialization.hpp"
+#include "src/Accelerators/PIM/Common/IR/WeightUtils.hpp"
+#include "src/Dialect/ONNX/ONNXOps.hpp"
+
+using namespace mlir;
+
+namespace onnx_mlir {
+
+bool isWeightLikeComputeOperand(Value value) {
+  auto rankedType = dyn_cast<RankedTensorType>(value.getType());
+  if (!rankedType || !isMatrixShape(rankedType.getShape()))
+    return false;
+
+  llvm::SmallPtrSet<Operation*, 8> visited;
+
+  while (auto* definingOp = value.getDefiningOp()) {
+    if (!visited.insert(definingOp).second)
+      return false;
+    if (isa<arith::ConstantOp, ONNXConstantOp>(definingOp) || hasWeightAlways(definingOp))
+      return true;
+
+    if (auto extractSliceOp = dyn_cast<tensor::ExtractSliceOp>(definingOp)) {
+      value = extractSliceOp.getSource();
+      continue;
+    }
+    if (auto expandShapeOp = dyn_cast<tensor::ExpandShapeOp>(definingOp)) {
+      value = expandShapeOp.getSrc();
+      continue;
+    }
+    if (auto collapseShapeOp = dyn_cast<tensor::CollapseShapeOp>(definingOp)) {
+      value = collapseShapeOp.getSrc();
+      continue;
+    }
+    if (auto transposeOp = dyn_cast<ONNXTransposeOp>(definingOp)) {
+      value = transposeOp.getData();
+      continue;
+    }
+
+    return false;
+  }
+
+  return false;
+}
+
+FailureOr<Value> materializeWeightLikeValueInBlock(Value value, IRRewriter& rewriter, IRMapping& mapper) {
+  if (auto mapped = mapper.lookupOrNull(value))
+    return cast<Value>(mapped);
+
+  Operation* definingOp = value.getDefiningOp();
+  if (!definingOp)
+    return failure();
+
+  if (isa<arith::ConstantOp, ONNXConstantOp>(definingOp)) {
+    auto tensorType = dyn_cast<RankedTensorType>(value.getType());
+    if (!tensorType || !tensorType.hasStaticShape())
+      return failure();
+
+    SmallVector<OpFoldResult> offsets(tensorType.getRank(), rewriter.getIndexAttr(0));
+    SmallVector<OpFoldResult> sizes;
+    SmallVector<OpFoldResult> strides(tensorType.getRank(), rewriter.getIndexAttr(1));
+    sizes.reserve(tensorType.getRank());
+    for (int64_t dim : tensorType.getShape())
+      sizes.push_back(rewriter.getIndexAttr(dim));
+
+    auto referencedValue =
+      tensor::ExtractSliceOp::create(rewriter, value.getLoc(), tensorType, value, offsets, sizes, strides);
+    mapper.map(value, referencedValue.getResult());
+    return referencedValue.getResult();
+  }
+
+  if (!isa<tensor::ExtractSliceOp, tensor::ExpandShapeOp, tensor::CollapseShapeOp, ONNXTransposeOp>(definingOp))
+    return failure();
+
+  IRMapping localMapper;
+  for (Value operand : definingOp->getOperands()) {
+    if (auto mapped = mapper.lookupOrNull(operand)) {
+      localMapper.map(operand, cast<Value>(mapped));
+      continue;
+    }
+
+    if (isWeightLikeComputeOperand(operand)) {
+      auto clonedOperand = materializeWeightLikeValueInBlock(operand, rewriter, mapper);
+      if (failed(clonedOperand))
+        return failure();
+      localMapper.map(operand, *clonedOperand);
+      continue;
+    }
+
+    localMapper.map(operand, operand);
+  }
+
+  Operation* clonedOp = rewriter.clone(*definingOp, localMapper);
+  for (auto [oldResult, newResult] : llvm::zip(definingOp->getResults(), clonedOp->getResults()))
+    mapper.map(oldResult, newResult);
+
+  auto mapped = mapper.lookupOrNull(value);
+  if (!mapped)
+    return failure();
+  return cast<Value>(mapped);
+}
+
+} // namespace onnx_mlir

src/PIM/Conversion/ONNXToSpatial/Common/WeightMaterialization.hpp

@@ -0,0 +1,18 @@
+#pragma once
+
+#include "mlir/IR/IRMapping.h"
+#include "mlir/IR/PatternMatch.h"
+#include "mlir/IR/Value.h"
+
+namespace onnx_mlir {
+
+/// Returns true when a matrix-valued compute operand is ultimately backed by a
+/// weight-marked constant/view chain and can be promoted into weights.
+bool isWeightLikeComputeOperand(mlir::Value value);
+
+/// Rebuilds the view/transpose chain of a promoted weight operand inside a new
+/// compute body while reusing already-materialized intermediate values.
+llvm::FailureOr<mlir::Value>
+materializeWeightLikeValueInBlock(mlir::Value value, mlir::IRRewriter& rewriter, mlir::IRMapping& mapper);
+
+} // namespace onnx_mlir

src/PIM/Conversion/ONNXToSpatial/ConversionPatterns.cpp

@@ -0,0 +1,32 @@
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Common/Common.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/ConversionPatterns.hpp"
+
+using namespace mlir;
+
+namespace onnx_mlir {
+
+namespace {
+
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/ONNXToSpatial.hpp.inc"
+
+} // namespace
+
+void populateConversionPatterns(mlir::RewritePatternSet& patterns, mlir::MLIRContext* ctx) {
+  patterns.add<removeLRN>(ctx);
+
+  populateElementwisePatterns(patterns, ctx);
+  populateGemmPatterns(patterns, ctx);
+  populateConvPatterns(patterns, ctx);
+  populatePoolPatterns(patterns, ctx);
+  populateReduceMeanPatterns(patterns, ctx);
+  populateReluPatterns(patterns, ctx);
+  populateSigmoidPatterns(patterns, ctx);
+  populateSoftmaxPatterns(patterns, ctx);
+  populateConcatPatterns(patterns, ctx);
+  populateGatherPatterns(patterns, ctx);
+  populateResizePatterns(patterns, ctx);
+  populateReshapePatterns(patterns, ctx);
+  populateSplitPatterns(patterns, ctx);
+}
+
+} // namespace onnx_mlir

src/PIM/Conversion/ONNXToSpatial/ConversionPatterns.hpp

@@ -5,6 +5,8 @@
 
 namespace onnx_mlir {
 
+void populateConversionPatterns(mlir::RewritePatternSet& patterns, mlir::MLIRContext* ctx);
+
 void populateConvPatterns(mlir::RewritePatternSet& patterns, mlir::MLIRContext* ctx);
 
 void populateElementwisePatterns(mlir::RewritePatternSet& patterns, mlir::MLIRContext* ctx);

src/PIM/Conversion/ONNXToSpatial/HostFoldability.cpp

@@ -0,0 +1,75 @@
+#include "mlir/Dialect/Arith/IR/Arith.h"
+#include "mlir/Dialect/Tensor/IR/Tensor.h"
+#include "mlir/IR/BuiltinTypes.h"
+
+#include "llvm/ADT/SmallPtrSet.h"
+
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/HostFoldability.hpp"
+#include "src/Accelerators/PIM/Dialect/Spatial/SpatialOps.hpp"
+#include "src/Dialect/ONNX/ONNXOps.hpp"
+
+using namespace mlir;
+
+namespace onnx_mlir {
+
+namespace {
+
+static bool hasStaticUnitStrides(tensor::ExtractSliceOp extractSliceOp) {
+  return llvm::all_of(extractSliceOp.getStaticStrides(), [](int64_t stride) { return stride == 1; });
+}
+
+static bool isStaticTensorResult(Operation* op) {
+  return llvm::all_of(op->getResultTypes(), [](Type type) {
+    auto shapedType = dyn_cast<ShapedType>(type);
+    return shapedType && shapedType.hasStaticShape();
+  });
+}
+
+static bool isHostFoldableOpImpl(Operation* op, llvm::SmallPtrSetImpl<Operation*>& visited) {
+  if (!op || !visited.insert(op).second)
+    return false;
+
+  if (isa<arith::ConstantOp, ONNXConstantOp, ONNXNoneOp>(op))
+    return true;
+
+  if (!isStaticTensorResult(op))
+    return false;
+
+  if (auto transposeOp = dyn_cast<ONNXTransposeOp>(op))
+    return isHostFoldableValue(transposeOp.getData());
+
+  if (auto collapseShapeOp = dyn_cast<tensor::CollapseShapeOp>(op))
+    return isHostFoldableValue(collapseShapeOp.getSrc());
+
+  if (auto expandShapeOp = dyn_cast<tensor::ExpandShapeOp>(op))
+    return isHostFoldableValue(expandShapeOp.getSrc());
+
+  if (auto extractSliceOp = dyn_cast<tensor::ExtractSliceOp>(op))
+    return hasStaticUnitStrides(extractSliceOp) && isHostFoldableValue(extractSliceOp.getSource());
+
+  if (auto extractRowsOp = dyn_cast<spatial::SpatExtractRowsOp>(op))
+    return isHostFoldableValue(extractRowsOp.getInput());
+
+  if (auto concatOp = dyn_cast<spatial::SpatConcatOp>(op))
+    return llvm::all_of(concatOp.getInputs(), isHostFoldableValue);
+
+  return false;
+}
+
+} // namespace
+
+bool isHostFoldableValue(Value value) {
+  auto* definingOp = value.getDefiningOp();
+  if (!definingOp)
+    return false;
+
+  llvm::SmallPtrSet<Operation*, 8> visited;
+  return isHostFoldableOpImpl(definingOp, visited);
+}
+
+bool isHostFoldableOp(Operation* op) {
+  llvm::SmallPtrSet<Operation*, 8> visited;
+  return isHostFoldableOpImpl(op, visited);
+}
+
+} // namespace onnx_mlir

src/PIM/Conversion/ONNXToSpatial/HostFoldability.hpp

@@ -0,0 +1,12 @@
+#pragma once
+
+#include "mlir/IR/Operation.h"
+#include "mlir/IR/Value.h"
+
+namespace onnx_mlir {
+
+bool isHostFoldableValue(mlir::Value value);
+
+bool isHostFoldableOp(mlir::Operation* op);
+
+} // namespace onnx_mlir

src/PIM/Conversion/ONNXToSpatial/HostLegality.cpp

@@ -0,0 +1,29 @@
+#include "mlir/Dialect/Arith/IR/Arith.h"
+#include "mlir/Dialect/Func/IR/FuncOps.h"
+#include "mlir/Dialect/Tensor/IR/Tensor.h"
+
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/HostFoldability.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/HostLegality.hpp"
+#include "src/Accelerators/PIM/Dialect/Spatial/SpatialOps.hpp"
+
+using namespace mlir;
+
+namespace onnx_mlir {
+
+LogicalResult verifyONNXToSpatialHostLegality(func::FuncOp funcOp) {
+  bool hasFailure = false;
+
+  for (Operation& op : funcOp.getFunctionBody().front()) {
+    if (isa<func::ReturnOp, spatial::SpatCompute, spatial::SpatComputeBatch>(&op))
+      continue;
+    if (isHostFoldableOp(&op))
+      continue;
+
+    op.emitOpError("non-foldable top-level runtime op remains after ONNX-to-Spatial; lower it inside spat.compute");
+    hasFailure = true;
+  }
+
+  return success(!hasFailure);
+}
+
+} // namespace onnx_mlir

src/PIM/Conversion/ONNXToSpatial/HostLegality.hpp

@@ -0,0 +1,10 @@
+#pragma once
+
+#include "mlir/Dialect/Func/IR/FuncOps.h"
+#include "mlir/Support/LogicalResult.h"
+
+namespace onnx_mlir {
+
+mlir::LogicalResult verifyONNXToSpatialHostLegality(mlir::func::FuncOp funcOp);
+
+} // namespace onnx_mlir

src/PIM/Conversion/ONNXToSpatial/ONNXToSpatialPass.cpp

@@ -1,3 +1,4 @@
+#include "mlir/Dialect/Arith/IR/Arith.h"
 #include "mlir/Dialect/Func/IR/FuncOps.h"
 #include "mlir/Dialect/SCF/IR/SCF.h"
 #include "mlir/Dialect/Tensor/IR/Tensor.h"
@@ -7,25 +8,18 @@
 #include "mlir/Transforms/GreedyPatternRewriteDriver.h"
 #include "mlir/Transforms/Passes.h"
 
-#include "llvm/ADT/STLExtras.h"
-#include "llvm/ADT/SmallSet.h"
 #include "llvm/ADT/SmallVector.h"
-#include "llvm/Support/Casting.h"
 #include "llvm/Support/Debug.h"
-#include "llvm/Support/raw_os_ostream.h"
 
-#include <fstream>
+#include "Common/Common.hpp"
-#include <iterator>
-#include <utility>
-
-#include "Common.hpp"
 #include "Common/PimCommon.hpp"
+#include "src/Accelerators/PIM/Common/PimCommon.hpp"
 #include "src/Accelerators/PIM/Compiler/PimCompilerOptions.hpp"
-#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Patterns.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/ConversionPatterns.hpp"
-#include "src/Accelerators/PIM/Dialect/Pim/PimOps.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/HostLegality.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/PostPatterns.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/PrePatterns.hpp"
 #include "src/Accelerators/PIM/Dialect/Spatial/SpatialOps.hpp"
-#include "src/Accelerators/PIM/Dialect/Spatial/Transforms/MergeComputeNodes/DCPGraph/DCPAnalysis.hpp"
-#include "src/Accelerators/PIM/Pass/PIMPasses.h"
 #include "src/Compiler/CompilerOptions.hpp"
 #include "src/Dialect/ONNX/ONNXOps.hpp"
 
@@ -33,12 +27,8 @@ using namespace mlir;
 
 namespace onnx_mlir {
 
-bool haveSameStaticShape(Value lhs, Value rhs);
-
 namespace {
 
-#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/ONNXToSpatial.hpp.inc"
-
 struct ONNXToSpatialPass : PassWrapper<ONNXToSpatialPass, OperationPass<ModuleOp>> {
   MLIR_DEFINE_EXPLICIT_INTERNAL_INLINE_TYPE_ID(ONNXToSpatialPass)
   StringRef getArgument() const override { return "convert-onnx-to-spatial"; }
@@ -48,33 +38,64 @@ struct ONNXToSpatialPass : PassWrapper<ONNXToSpatialPass, OperationPass<ModuleOp
   ONNXToSpatialPass(const ONNXToSpatialPass& pass) {}
 
   void runOnOperation() override;
-
-private:
-  void annotateWeightsConstants(func::FuncOp funcOp) const;
-  void encapsulateGlobalInstruction(func::FuncOp funcOp);
-  void mergeTriviallyConnectedComputes(func::FuncOp funcOp);
-  LogicalResult promoteConstantInputsToWeights(func::FuncOp funcOp);
 };
 
 } // namespace
 
+static void populateEmptyFunction(func::FuncOp funcOp) {
+  IRRewriter rewriter(funcOp.getContext());
+  IRMapping mapper;
+  SmallVector<spatial::SpatCompute> computes(funcOp.getOps<spatial::SpatCompute>());
+  if (!computes.empty())
+    return;
+
+  auto returnOp = cast<func::ReturnOp>(funcOp.getFunctionBody().front().getTerminator());
+  rewriter.setInsertionPoint(returnOp);
+
+  SmallVector<Type> sourceTypes;
+  SmallVector<Location> sourceLocs;
+  sourceTypes.reserve(funcOp.getNumArguments());
+  sourceLocs.reserve(funcOp.getNumArguments());
+  for (Value source : funcOp.getArguments()) {
+    sourceTypes.push_back(source.getType());
+    sourceLocs.push_back(source.getLoc());
+  }
+
+  auto newCompute = spatial::SpatCompute::create(
+    rewriter, returnOp.getLoc(), returnOp.getOperandTypes(), funcOp.getArguments(), {}, {});
+  auto* newBlock = rewriter.createBlock(&newCompute.getBody(), newCompute.getBody().end(), sourceTypes, sourceLocs);
+  for (auto [blockArg, computeArg] : llvm::zip(newBlock->getArguments(), newCompute.getOperands()))
+    mapper.map(computeArg, blockArg);
+  newCompute.getProperties().setOperandSegmentSizes({0, static_cast<int>(sourceTypes.size())});
+
+  rewriter.setInsertionPointToEnd(newBlock);
+  for (Operation& op : funcOp.getOps())
+    if (!isa<spatial::SpatCompute, func::ReturnOp>(&op))
+      rewriter.clone(op, mapper);
+
+  auto yield = spatial::SpatYieldOp::create(rewriter, funcOp.getLoc(), returnOp.getOperands());
+  for (size_t i = 0; i < yield.getNumOperands(); ++i)
+    yield.setOperand(i, mapper.lookupOrDefault(yield.getOperand(i)));
+
+  for (Operation& op : llvm::make_early_inc_range(funcOp.getOps()))
+    if (!isa<spatial::SpatCompute, func::ReturnOp>(&op)) {
+      op.dropAllUses();
+      rewriter.eraseOp(&op);
+    }
+
+  for (auto [index, computeResult] : llvm::enumerate(newCompute.getResults()))
+    returnOp.setOperand(index, computeResult);
+}
+
 void ONNXToSpatialPass::runOnOperation() {
   ModuleOp moduleOp = getOperation();
   MLIRContext* ctx = &getContext();
 
-  RewritePatternSet mergeActivationPatterns(ctx);
+  RewritePatternSet prePatterns(ctx);
-  mergeActivationPatterns.add<onnxToArithConstant>(ctx);
+  populatePrePatterns(prePatterns, ctx);
-  mergeActivationPatterns.add<convAddToConvWithBiasLeft>(ctx);
+  if (failed(applyPatternsGreedily(moduleOp, std::move(prePatterns))))
-  mergeActivationPatterns.add<convAddToConvWithBiasRight>(ctx);
+    llvm::dbgs() << "Failed to apply pre-patterns, continuing...\n";
-  mergeActivationPatterns.add<matMulAddToGemm>(ctx);
-  mergeActivationPatterns.add<matMulToGemm>(ctx);
-  mergeActivationPatterns.add<removeFlattenSameShape>(ctx);
-  populateMatMulRewritePatterns(mergeActivationPatterns, ctx);
 
-  if (failed(applyPatternsGreedily(moduleOp, std::move(mergeActivationPatterns))))
-    llvm::dbgs() << "Failed to merge activation patterns, continuing...\n";
-
-  IRRewriter rewriter(moduleOp);
   auto entryFunc = getPimEntryFunc(moduleOp);
   if (failed(entryFunc)) {
     signalPassFailure();
@@ -87,8 +108,7 @@ void ONNXToSpatialPass::runOnOperation() {
                          tensor::TensorDialect,
                          arith::ArithDialect,
                          scf::SCFDialect>();
-  target.addDynamicallyLegalOp<ONNXMatMulOp>(
+  target.addIllegalOp<ONNXMatMulOp>();
-    [](ONNXMatMulOp op) { return cast<ShapedType>(op.getY().getType()).getRank() != 2; });
   target.addIllegalOp<ONNXAddOp>();
   target.addIllegalOp<ONNXDivOp>();
   target.addIllegalOp<ONNXMulOp>();
@@ -107,32 +127,23 @@ void ONNXToSpatialPass::runOnOperation() {
   target.addIllegalOp<ONNXReduceMeanV13Op>();
   target.addIllegalOp<ONNXSplitOp>();
 
-  RewritePatternSet patterns(ctx);
+  RewritePatternSet conversionPatterns(ctx);
-  patterns.add<removeLRN>(ctx);
+  populateConversionPatterns(conversionPatterns, ctx);
-
+  if (failed(applyPartialConversion(moduleOp, target, std::move(conversionPatterns)))) {
-  populateElementwisePatterns(patterns, ctx);
+    signalPassFailure();
-  populateGemmPatterns(patterns, ctx);
+    return;
-  populateConvPatterns(patterns, ctx);
+  }
-  populatePoolPatterns(patterns, ctx);
+
-  populateReduceMeanPatterns(patterns, ctx);
+  RewritePatternSet earlyPostPatterns(ctx);
-  populateReluPatterns(patterns, ctx);
+  populateEarlyPostPatterns(earlyPostPatterns, ctx);
-  populateSigmoidPatterns(patterns, ctx);
+  if (failed(applyPatternsGreedily(*entryFunc, std::move(earlyPostPatterns)))) {
-  populateSoftmaxPatterns(patterns, ctx);
-  populateConcatPatterns(patterns, ctx);
-  populateGatherPatterns(patterns, ctx);
-  populateResizePatterns(patterns, ctx);
-  populateReshapePatterns(patterns, ctx);
-  populateSplitPatterns(patterns, ctx);
-
-  if (failed(applyPartialConversion(moduleOp, target, std::move(patterns)))) {
     signalPassFailure();
     return;
   }
 
-  // Count the number of compute ops and check they do not exceed the core count
   if (coresCount != -1) {
     int computeOpsCount = 0;
-    for (auto& op : entryFunc->getFunctionBody().front().getOperations())
+    for (Operation& op : entryFunc->getFunctionBody().front().getOperations())
       if (isa<spatial::SpatCompute>(op))
         computeOpsCount++;
 
@@ -149,334 +160,24 @@ void ONNXToSpatialPass::runOnOperation() {
     llvm::dbgs() << "Failed to run canonicalization cleanup, continuing...\n";
 
   annotateWeightsConstants(*entryFunc);
-  encapsulateGlobalInstruction(*entryFunc);
 
-  if (failed(promoteConstantInputsToWeights(*entryFunc))) {
+  RewritePatternSet postPatterns(ctx);
+  populatePostPatterns(postPatterns, ctx);
+  if (failed(applyPatternsGreedily(*entryFunc, std::move(postPatterns)))) {
     signalPassFailure();
     return;
   }
 
-  mergeTriviallyConnectedComputes(*entryFunc);
+  if (failed(verifyONNXToSpatialHostLegality(*entryFunc))) {
+    signalPassFailure();
+    return;
+  }
+
+  populateEmptyFunction(*entryFunc);
 
-  // Dump to file for debug
   dumpModule(moduleOp, "spatial0");
 }
 
-template <typename T>
-bool encapsulator(IRRewriter& rewriter, Location loc, Operation* inst, std::function<Value(T)> funcSource) {
-  if (T toRemoveOp = llvm::dyn_cast_if_present<T>(inst)) {
-    Value source = funcSource(toRemoveOp);
-    rewriter.setInsertionPointAfter(toRemoveOp);
-    if (isa_and_present<spatial::SpatCompute>(source.getDefiningOp())) {
-      auto newCompute = spatial::SpatCompute::create(rewriter, loc, inst->getResultTypes(), source);
-      auto BB = rewriter.createBlock(&newCompute.getBody(), newCompute.getBody().end(), {source.getType()}, {loc});
-      newCompute.getProperties().setOperandSegmentSizes({(int) 0, (int) 1});
-      rewriter.setInsertionPointToEnd(BB);
-      IRMapping mapper;
-      mapper.map(source, BB->getArgument(0));
-      auto newInst = rewriter.clone(*inst, mapper);
-      spatial::SpatYieldOp::create(rewriter, loc, newInst->getResults());
-      inst->replaceAllUsesWith(newCompute->getResults());
-      inst->erase();
-      return true;
-    }
-  }
-  return false;
-}
-
-bool encapsulateConcat(IRRewriter& rewriter, Location loc, Operation* inst) {
-  if (auto toRemoveOp = llvm::dyn_cast_if_present<tensor::ConcatOp>(inst)) {
-    auto sources = toRemoveOp.getInputs();
-    rewriter.setInsertionPointAfter(toRemoveOp);
-    if (llvm::any_of(
-          sources, [](auto source) { return isa_and_present<spatial::SpatCompute>(source.getDefiningOp()); })) {
-      auto newCompute = spatial::SpatCompute::create(rewriter, loc, inst->getResultTypes(), sources);
-      SmallVector<Type> sourceTypes;
-      SmallVector<Location> sourceLoc;
-      for (auto source : sources) {
-        sourceTypes.push_back(source.getType());
-        sourceLoc.push_back(loc);
-      }
-      auto BB = rewriter.createBlock(&newCompute.getBody(), newCompute.getBody().end(), sourceTypes, sourceLoc);
-      newCompute.getProperties().setOperandSegmentSizes({(int) 0, (int) sources.size()});
-      rewriter.setInsertionPointToEnd(BB);
-      IRMapping mapper;
-      for (auto [source, bbArg] : llvm::zip(sources, BB->getArguments()))
-        mapper.map(source, bbArg);
-      auto newConcat = rewriter.clone(*inst, mapper);
-      spatial::SpatYieldOp::create(rewriter, loc, newConcat->getResults());
-      inst->replaceAllUsesWith(newCompute->getResults());
-      inst->erase();
-      return true;
-    }
-  }
-  return false;
-}
-
-static FailureOr<Value> materializeWeightLikeValueInBlock(Value value, IRRewriter& rewriter, IRMapping& mapper) {
-  if (auto mapped = mapper.lookupOrNull(value))
-    return cast<Value>(mapped);
-
-  Operation* definingOp = value.getDefiningOp();
-  if (!definingOp)
-    return failure();
-
-  if (isa<arith::ConstantOp, ONNXConstantOp>(definingOp)) {
-    auto tensorType = dyn_cast<RankedTensorType>(value.getType());
-    if (!tensorType || !tensorType.hasStaticShape())
-      return failure();
-
-    SmallVector<OpFoldResult> offsets(tensorType.getRank(), rewriter.getIndexAttr(0));
-    SmallVector<OpFoldResult> sizes;
-    SmallVector<OpFoldResult> strides(tensorType.getRank(), rewriter.getIndexAttr(1));
-    sizes.reserve(tensorType.getRank());
-    for (int64_t dim : tensorType.getShape())
-      sizes.push_back(rewriter.getIndexAttr(dim));
-
-    auto referencedValue =
-      tensor::ExtractSliceOp::create(rewriter, value.getLoc(), tensorType, value, offsets, sizes, strides);
-    mapper.map(value, referencedValue.getResult());
-    return referencedValue.getResult();
-  }
-
-  if (!isa<tensor::ExtractSliceOp, tensor::ExpandShapeOp, tensor::CollapseShapeOp, ONNXTransposeOp>(definingOp))
-    return failure();
-
-  IRMapping localMapper;
-  for (Value operand : definingOp->getOperands()) {
-    if (auto mapped = mapper.lookupOrNull(operand)) {
-      localMapper.map(operand, cast<Value>(mapped));
-      continue;
-    }
-
-    if (isWeightLikeComputeOperand(operand)) {
-      auto clonedOperand = materializeWeightLikeValueInBlock(operand, rewriter, mapper);
-      if (failed(clonedOperand))
-        return failure();
-      localMapper.map(operand, *clonedOperand);
-      continue;
-    }
-
-    localMapper.map(operand, operand);
-  }
-
-  Operation* clonedOp = rewriter.clone(*definingOp, localMapper);
-  for (auto [oldResult, newResult] : llvm::zip(definingOp->getResults(), clonedOp->getResults()))
-    mapper.map(oldResult, newResult);
-
-  auto mapped = mapper.lookupOrNull(value);
-  if (!mapped)
-    return failure();
-  return cast<Value>(mapped);
-}
-
-// TODO what we want to keep in global?
-void ONNXToSpatialPass::encapsulateGlobalInstruction(func::FuncOp funcOp) {
-  Location loc = funcOp.getLoc();
-  IRRewriter rewriter(&getContext());
-  bool keep = true;
-  while (keep) {
-    keep = false;
-    for (auto& instruction : llvm::make_early_inc_range(funcOp.getOps())) {
-      keep |= encapsulator<tensor::ExtractSliceOp>(
-        rewriter, loc, &instruction, [](tensor::ExtractSliceOp extract) { return extract.getSource(); });
-
-      keep |= encapsulator<tensor::ExpandShapeOp>(
-        rewriter, loc, &instruction, [](tensor::ExpandShapeOp expand) { return expand.getSrc(); });
-
-      keep |= encapsulator<ONNXTransposeOp>(
-        rewriter, loc, &instruction, [](ONNXTransposeOp transpose) { return transpose.getData(); });
-
-      keep |= encapsulator<tensor::CollapseShapeOp>(
-        rewriter, loc, &instruction, [](tensor::CollapseShapeOp collapse) { return collapse.getSrc(); });
-
-      keep |= encapsulateConcat(rewriter, loc, &instruction);
-    }
-  }
-}
-
-void ONNXToSpatialPass::mergeTriviallyConnectedComputes(func::FuncOp funcOp) {
-  Location loc = funcOp.getLoc();
-  IRRewriter rewriter(&getContext());
-  SmallVector<spatial::SpatCompute> trivialComputes;
-  llvm::SmallSet<spatial::SpatCompute, 8> toErase;
-
-  for (auto compute : funcOp.getOps<spatial::SpatCompute>())
-    if (compute->hasOneUse()) {
-      auto& use = *compute->getUses().begin();
-      auto user = dyn_cast<spatial::SpatCompute>(use.getOwner());
-
-      if (user && user.getInputs().size() == 1 && use.getOperandNumber() >= user.getWeights().size())
-        trivialComputes.push_back(compute);
-    }
-
-  while (!trivialComputes.empty()) {
-    auto compute = trivialComputes.front();
-
-    if (compute.use_empty()) {
-      std::swap(trivialComputes.front(), trivialComputes.back());
-      trivialComputes.pop_back();
-      continue;
-    }
-    auto& computeUse = *compute->getUses().begin();
-    auto child = cast<spatial::SpatCompute>(computeUse.getOwner());
-    auto usedResult = cast<OpResult>(computeUse.get()).getResultNumber();
-    auto childArgIndex = computeUse.getOperandNumber() - child.getWeights().size();
-
-    rewriter.setInsertionPointAfter(compute.getOperation());
-
-    auto newCompute =
-      spatial::SpatCompute::create(rewriter, loc, child.getResultTypes(), compute.getOperands());
-    newCompute.getProperties().setOperandSegmentSizes(
-      {static_cast<int>(compute.getWeights().size()), static_cast<int>(compute.getInputs().size())});
-
-    IRMapping mapper;
-    auto weightMutableIter = newCompute.getWeightsMutable();
-    for (auto weight : child.getWeights()) {
-      auto founded = llvm::find(newCompute.getWeights(), weight);
-      if (founded == newCompute.getWeights().end()) {
-        weightMutableIter.append(weight);
-        auto last = weightMutableIter.end();
-        last = std::prev(last, 1);
-        mapper.map(weight, last->get());
-      }
-      else {
-        mapper.map(weight, *founded);
-      }
-    }
-
-    compute.getBodyRegion().cloneInto(&newCompute.getBodyRegion(), mapper);
-    auto newTerminator = newCompute.getBody().front().getTerminator();
-    mapper.map(child.getBody().front().getArgument(childArgIndex), newTerminator->getOperand(usedResult));
-    newTerminator->erase();
-    rewriter.setInsertionPoint(&newCompute.getBody().front(), newCompute.getBody().front().end());
-    for (auto& op : child.getBody().front()) {
-      auto newInst = rewriter.clone(op, mapper);
-
-      if (auto vmOp = llvm::dyn_cast<spatial::SpatWeightedMVMOp>(newInst)) {
-        auto oldIndex = vmOp.getWeightIndex();
-        auto newWeight = mapper.lookup(*std::next(child.getWeights().begin(), oldIndex));
-        auto newIndex = std::distance(newCompute.getWeights().begin(), llvm::find(newCompute.getWeights(), newWeight));
-        vmOp.setWeightIndex(newIndex);
-      }
-      if (auto vmOp = llvm::dyn_cast<spatial::SpatWeightedVMMOp>(newInst)) {
-        auto oldIndex = vmOp.getWeightIndex();
-        auto newWeight = mapper.lookup(*std::next(child.getWeights().begin(), oldIndex));
-        auto newIndex = std::distance(newCompute.getWeights().begin(), llvm::find(newCompute.getWeights(), newWeight));
-        vmOp.setWeightIndex(newIndex);
-      }
-    }
-
-    child.replaceAllUsesWith(newCompute);
-    toErase.insert(child);
-
-    std::swap(trivialComputes.front(), trivialComputes.back());
-    trivialComputes.pop_back();
-    toErase.insert(compute);
-
-    if (newCompute->hasOneUse()) {
-      auto& use = *newCompute->getUses().begin();
-      auto user = dyn_cast<spatial::SpatCompute>(use.getOwner());
-      if (user && user.getInputs().size() == 1 && use.getOperandNumber() >= user.getWeights().size())
-        trivialComputes.push_back(newCompute);
-    }
-  }
-
-  for (auto compute : toErase) {
-    for (Value result : compute->getResults())
-      result.dropAllUses();
-    compute.erase();
-  }
-}
-
-void ONNXToSpatialPass::annotateWeightsConstants(func::FuncOp funcOp) const {
-  funcOp.walk([&](arith::ConstantOp constantOp) {
-    bool isAlwaysWeight =
-      llvm::all_of(constantOp->getUsers(), [](auto user) -> bool { return isa<spatial::SpatCompute>(user); });
-    if (isAlwaysWeight)
-      markWeightAlways(constantOp);
-  });
-}
-
-LogicalResult ONNXToSpatialPass::promoteConstantInputsToWeights(func::FuncOp funcOp) {
-  IRRewriter rewriter(&getContext());
-  SmallVector<spatial::SpatCompute> computes(funcOp.getOps<spatial::SpatCompute>());
-
-  for (auto compute : computes) {
-    SmallVector<bool> promoteInput(compute.getInputs().size(), false);
-    bool needsRewrite = false;
-    for (auto [inputIdx, input] : llvm::enumerate(compute.getInputs())) {
-      if (!isWeightLikeComputeOperand(input))
-        continue;
-      promoteInput[inputIdx] = true;
-      needsRewrite = true;
-    }
-    if (!needsRewrite)
-      continue;
-
-    rewriter.setInsertionPointAfter(compute);
-
-    SmallVector<Value> newWeights(compute.getWeights().begin(), compute.getWeights().end());
-    SmallVector<Value> newInputs;
-    SmallVector<Type> newInputTypes;
-    SmallVector<Location> newInputLocs;
-    newWeights.reserve(compute.getWeights().size() + compute.getInputs().size());
-    newInputs.reserve(compute.getInputs().size());
-    newInputTypes.reserve(compute.getInputs().size());
-    newInputLocs.reserve(compute.getInputs().size());
-
-    for (auto [inputIdx, input] : llvm::enumerate(compute.getInputs())) {
-      if (promoteInput[inputIdx]) {
-        newWeights.push_back(input);
-        continue;
-      }
-      newInputs.push_back(input);
-      newInputTypes.push_back(input.getType());
-      newInputLocs.push_back(input.getLoc());
-    }
-
-    auto newCompute =
-      spatial::SpatCompute::create(rewriter, compute.getLoc(), compute.getResultTypes(), newWeights, newInputs);
-    auto* newBlock =
-      rewriter.createBlock(&newCompute.getBody(), newCompute.getBody().end(), newInputTypes, newInputLocs);
-    newCompute.getProperties().setOperandSegmentSizes(
-      {static_cast<int>(newWeights.size()), static_cast<int>(newInputs.size())});
-    rewriter.setInsertionPointToStart(newBlock);
-
-    IRMapping mapper;
-    auto& oldBlock = compute.getBody().front();
-    size_t newInputIdx = 0;
-    for (auto [oldInputIdx, oldArg] : llvm::enumerate(oldBlock.getArguments())) {
-      if (!promoteInput[oldInputIdx]) {
-        mapper.map(oldArg, newBlock->getArgument(newInputIdx++));
-        continue;
-      }
-
-      auto clonedValue = materializeWeightLikeValueInBlock(compute.getInputs()[oldInputIdx], rewriter, mapper);
-      if (failed(clonedValue))
-        return compute.emitError("failed to materialize promoted weight-like operand inside compute body");
-      mapper.map(oldArg, *clonedValue);
-    }
-
-    for (auto& op : oldBlock.without_terminator())
-      rewriter.clone(op, mapper);
-
-    auto oldYield = cast<spatial::SpatYieldOp>(oldBlock.getTerminator());
-    SmallVector<Value> newYieldOperands;
-    newYieldOperands.reserve(oldYield.getOutputs().size());
-    for (Value operand : oldYield.getOutputs()) {
-      auto mapped = mapper.lookupOrNull(operand);
-      newYieldOperands.push_back(mapped ? cast<Value>(mapped) : operand);
-    }
-    spatial::SpatYieldOp::create(rewriter, oldYield.getLoc(), newYieldOperands);
-
-    compute.replaceAllUsesWith(newCompute);
-    compute.erase();
-  }
-
-  return success();
-}
-
 std::unique_ptr<Pass> createONNXToSpatialPass() { return std::make_unique<ONNXToSpatialPass>(); }
 
 } // namespace onnx_mlir

src/PIM/Conversion/ONNXToSpatial/Patterns/Math/Conv.cpp

@@ -7,11 +7,10 @@
 #include "llvm/ADT/SmallVector.h"
 
 #include <algorithm>
-#include <cassert>
 
-#include "src/Accelerators/PIM/Common/PimCommon.hpp"
+#include "src/Accelerators/PIM/Common/Support/Diagnostics.hpp"
 #include "src/Accelerators/PIM/Compiler/PimCompilerOptions.hpp"
-#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Common.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Common/Common.hpp"
 #include "src/Accelerators/PIM/Dialect/Spatial/SpatialOps.hpp"
 #include "src/Dialect/ONNX/ONNXOps.hpp"
 
@@ -147,11 +146,11 @@ static Value buildPackedBias(bool hasBias,
   return arith::ConstantOp::create(rewriter, loc, packedBiasType, packedBiasAttr).getResult();
 }
 
-static SmallVector<Value> createIm2colRowComputes(Value x,
+static Value createIm2colRowComputes(Value x,
                                      RankedTensorType xType,
                                      RankedTensorType im2colType,
                                      RankedTensorType im2colRowType,
-                                                  RankedTensorType gemmInputRowType,
+                                     RankedTensorType gemmInputRowsType,
                                      int64_t batchSize,
                                      int64_t numChannelsIn,
                                      int64_t xHeight,
@@ -176,8 +175,8 @@ static SmallVector<Value> createIm2colRowComputes(Value x,
   auto elemType = xType.getElementType();
   constexpr size_t numInputs = 1;
   const int64_t packedNumRows = ceilIntegerDivide(numPatches, packFactor);
-  SmallVector<Type> resultTypes(packedNumRows, gemmInputRowType);
+  auto im2colComputeOp =
-  auto im2colComputeOp = createSpatCompute<numInputs>(rewriter, loc, resultTypes, {}, x, [&](Value xArg) {
+    createSpatCompute<numInputs>(rewriter, loc, TypeRange {gemmInputRowsType}, {}, x, [&](Value xArg) {
       Value paddedInput = xArg;
 
       // Pad input with zeros if needed:
@@ -285,24 +284,10 @@ static SmallVector<Value> createIm2colRowComputes(Value x,
         });
       }
 
-    SmallVector<Value> rowResults;
+      spatial::SpatYieldOp::create(rewriter, loc, gemmInputRows);
-    rowResults.reserve(packedNumRows);
-    for (int64_t rowIdx = 0; rowIdx < packedNumRows; rowIdx++) {
-      SmallVector<OpFoldResult> offsets = {rewriter.getIndexAttr(rowIdx), rewriter.getIndexAttr(0)};
-      SmallVector<OpFoldResult> sizes = {rewriter.getIndexAttr(1),
-                                         rewriter.getIndexAttr(packFactor * patchSize)};
-      SmallVector<OpFoldResult> strides = {rewriter.getIndexAttr(1), rewriter.getIndexAttr(1)};
-      rowResults.push_back(
-        tensor::ExtractSliceOp::create(rewriter, loc, gemmInputRowType, gemmInputRows, offsets, sizes, strides));
-    }
-    spatial::SpatYieldOp::create(rewriter, loc, rowResults);
     });
 
-  SmallVector<Value> rows;
+  return im2colComputeOp.getResult(0);
-  rows.reserve(im2colComputeOp.getNumResults());
-  for (Value result : im2colComputeOp.getResults())
-    rows.push_back(result);
-  return rows;
 }
 
 static Value createCollectedConvOutput(ValueRange gemmRows,
@@ -320,16 +305,12 @@ static Value createCollectedConvOutput(ValueRange gemmRows,
   auto collectComputeOp = createSpatCompute(rewriter, loc, convType, {}, gemmRows, [&](ValueRange gemmRowArgs) {
     Value gemmOut;
     if (packFactor == 1) {
-      gemmOut = gemmRowArgs.size() == 1 ? gemmRowArgs.front()
+      gemmOut = createSpatConcat(rewriter, loc, /*axis=*/0, gemmRowArgs);
-                                        : tensor::ConcatOp::create(rewriter, loc, /*axis=*/0, gemmRowArgs).getResult();
     }
     else {
       auto expandedType = RankedTensorType::get({packedNumRows, packFactor, numChannelsOut}, outType.getElementType());
       auto paddedType = RankedTensorType::get({paddedNumPatches, numChannelsOut}, outType.getElementType());
-      Value packedOutput =
+      Value packedOutput = createSpatConcat(rewriter, loc, /*axis=*/0, gemmRowArgs);
-        gemmRowArgs.size() == 1
-          ? gemmRowArgs.front()
-          : tensor::ConcatOp::create(rewriter, loc, /*axis=*/0, gemmRowArgs).getResult();
       Value expandedOutput = tensor::ExpandShapeOp::create(rewriter,
                                                            loc,
                                                            expandedType,
@@ -388,11 +369,34 @@ LogicalResult ConvToGemm::matchAndRewrite(ONNXConvOp convOp,
   auto wType = cast<RankedTensorType>(w.getType());
   auto outType = cast<RankedTensorType>(convOp.getY().getType());
 
-  assert("Only support static shapes" && xType.hasStaticShape() && wType.hasStaticShape() && outType.hasStaticShape());
+  if (!xType.hasStaticShape()) {
-  assert("Only support 2D convolution" && xType.getRank() == 4);
+    pim::emitUnsupportedStaticShapeDiagnostic(convOp, "conv input");
-
+    return failure();
-  // We need to understand what is group
+  }
-  assert("Only support group=1" && convOp.getGroup() == 1);
+  if (!wType.hasStaticShape()) {
+    pim::emitUnsupportedStaticShapeDiagnostic(convOp, "conv weight");
+    return failure();
+  }
+  if (!outType.hasStaticShape()) {
+    pim::emitUnsupportedStaticShapeDiagnostic(convOp, "conv result");
+    return failure();
+  }
+  if (xType.getRank() != 4) {
+    pim::emitUnsupportedRankDiagnostic(convOp, "conv input", xType.getRank(), {4});
+    return failure();
+  }
+  if (wType.getRank() != 4) {
+    pim::emitUnsupportedRankDiagnostic(convOp, "conv weight", wType.getRank(), {4});
+    return failure();
+  }
+  if (outType.getRank() != 4) {
+    pim::emitUnsupportedRankDiagnostic(convOp, "conv result", outType.getRank(), {4});
+    return failure();
+  }
+  if (convOp.getGroup() != 1) {
+    convOp.emitOpError("only group=1 convolution is supported for Spatial lowering");
+    return failure();
+  }
 
   const int64_t batchSize = xType.getDimSize(0);
   const int64_t numChannelsIn = xType.getDimSize(1);
@@ -409,6 +413,19 @@ LogicalResult ConvToGemm::matchAndRewrite(ONNXConvOp convOp,
   const auto dilationsAttr = convOp.getDilations();
   const auto padsAttr = convOp.getPads();
 
+  if (stridesAttr && stridesAttr->size() != 2) {
+    convOp.emitOpError("requires exactly two stride values for Spatial lowering");
+    return failure();
+  }
+  if (dilationsAttr && dilationsAttr->size() != 2) {
+    convOp.emitOpError("requires exactly two dilation values for Spatial lowering");
+    return failure();
+  }
+  if (padsAttr && padsAttr->size() != 4) {
+    convOp.emitOpError("requires exactly four pad values for 2D Spatial lowering");
+    return failure();
+  }
+
   const int64_t strideHeight = stridesAttr ? getI64FromArrayAttr(*stridesAttr, 0) : 1;
   const int64_t strideWidth = stridesAttr ? getI64FromArrayAttr(*stridesAttr, 1) : 1;
   const int64_t dilationHeight = dilationsAttr ? getI64FromArrayAttr(*dilationsAttr, 0) : 1;
@@ -449,6 +466,10 @@ LogicalResult ConvToGemm::matchAndRewrite(ONNXConvOp convOp,
         padWidthBegin = totalPadW - padWidthEnd;
       }
     }
+    else if (autoPad != "NOTSET" && autoPad != "VALID") {
+      convOp.emitOpError() << "unsupported auto_pad value `" << autoPad << "` for Spatial lowering";
+      return failure();
+    }
     // "NOTSET" or "VALID" -> all pads stay 0
   }
 
@@ -505,17 +526,21 @@ LogicalResult ConvToGemm::matchAndRewrite(ONNXConvOp convOp,
   //   B (weights): [patchSize, cOut]        -- W^T, stored in crossbar columns
   // and optionally repack several old rows into one GEMM row to use the available crossbar size better.
   //
-  // The im2col compute yields each GEMM input row as a separate result so every GEMM consumes only
+  // We want to process N pixels at the same time. Instead of doing N separate operations
-  // the row it needs instead of receiving a full packed tensor and slicing it locally.
+  // of (1 x patchSize) x (patchSize x cOut), we construct a block-diagonal weight matrix
-  auto gemmInputRowType =
+  // containing N copies of W^T and concatenate N im2col rows into one longer row:
-    RankedTensorType::get({1, effectiveMaxParallelPixels * patchSize}, elemType);
+  //   A_packed: [ceil(numPatches / N), N * patchSize]
-  auto gemmOutputRowType =
+  //   B_packed: [N * patchSize, N * cOut]
-    RankedTensorType::get({1, effectiveMaxParallelPixels * numChannelsOut}, outType.getElementType());
+  //   Y_packed: [ceil(numPatches / N), N * cOut]
-  SmallVector<Value> gemmInputRows = createIm2colRowComputes(x,
+  const int64_t packedNumRows = ceilIntegerDivide(numPatches, effectiveMaxParallelPixels);
+  auto gemmInputRowsType = RankedTensorType::get({packedNumRows, effectiveMaxParallelPixels * patchSize}, elemType);
+  auto gemmOutputRowsType =
+    RankedTensorType::get({packedNumRows, effectiveMaxParallelPixels * numChannelsOut}, outType.getElementType());
+  Value gemmInputRows = createIm2colRowComputes(x,
                                                 xType,
                                                 im2colType,
                                                 rowType,
-                                                            gemmInputRowType,
+                                                gemmInputRowsType,
                                                 batchSize,
                                                 numChannelsIn,
                                                 xHeight,
@@ -552,13 +577,10 @@ LogicalResult ConvToGemm::matchAndRewrite(ONNXConvOp convOp,
   Value gemmC = buildPackedBias(
     hasB, gemmBias, biasMatrix, biasDenseAttr, outType, numChannelsOut, effectiveMaxParallelPixels, rewriter, loc);
 
-  SmallVector<Value> gemmRows;
+  Value gemmRows = ONNXGemmOp::create(rewriter,
-  gemmRows.reserve(gemmInputRows.size());
-  for (Value gemmInputRow : gemmInputRows) {
-    Value gemmRow = ONNXGemmOp::create(rewriter,
                                       loc,
-                                       gemmOutputRowType,
+                                      gemmOutputRowsType,
-                                       gemmInputRow,
+                                      gemmInputRows,
                                       gemmB,
                                       gemmC,
                                       rewriter.getF32FloatAttr(1.0f),
@@ -566,11 +588,9 @@ LogicalResult ConvToGemm::matchAndRewrite(ONNXConvOp convOp,
                                       rewriter.getBoolAttr(false),
                                       rewriter.getBoolAttr(false))
                      .getY();
-    gemmRows.push_back(gemmRow);
-  }
 
   rewriter.replaceOp(convOp,
-                     createCollectedConvOutput(gemmRows,
+                     createCollectedConvOutput(ValueRange {gemmRows},
                                                convOp.getType(),
                                                gemmOutType,
                                                nhwcType,

src/PIM/Conversion/ONNXToSpatial/Patterns/Math/Elementwise.cpp

@@ -5,8 +5,9 @@
 
 #include "llvm/ADT/SmallVector.h"
 
-#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Common.hpp"
+#include "src/Accelerators/PIM/Common/IR/ShapeUtils.hpp"
-#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Patterns.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Common/Common.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/ConversionPatterns.hpp"
 #include "src/Accelerators/PIM/Dialect/Spatial/SpatialOps.hpp"
 #include "src/Dialect/ONNX/ONNXOps.hpp"
 
@@ -15,13 +16,6 @@ using namespace mlir;
 namespace onnx_mlir {
 namespace {
 
-static SmallVector<int64_t> computeRowMajorStrides(ArrayRef<int64_t> shape) {
-  SmallVector<int64_t> strides(shape.size(), 1);
-  for (int64_t i = static_cast<int64_t>(shape.size()) - 2; i >= 0; --i)
-    strides[i] = strides[i + 1] * shape[i + 1];
-  return strides;
-}
-
 static DenseElementsAttr getDenseConstantAttr(Value value) {
   if (auto constantOp = value.getDefiningOp<arith::ConstantOp>())
     return dyn_cast<DenseElementsAttr>(constantOp.getValue());

src/PIM/Conversion/ONNXToSpatial/Patterns/Math/Gemm.cpp

@@ -1,16 +1,17 @@
 #include "mlir/Dialect/Tensor/IR/Tensor.h"
 #include "mlir/Dialect/Tosa/IR/TosaOps.h"
 #include "mlir/IR/BuiltinTypes.h"
+#include "mlir/IR/IRMapping.h"
 #include "mlir/IR/Location.h"
 #include "mlir/Support/LogicalResult.h"
 #include "mlir/Transforms/DialectConversion.h"
 
 #include "llvm/ADT/SmallVector.h"
 
-#include <cassert>
-
 #include "src/Accelerators/PIM/Common/PimCommon.hpp"
-#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Common.hpp"
+#include "src/Accelerators/PIM/Common/Support/Diagnostics.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Common/Common.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/HostFoldability.hpp"
 #include "src/Accelerators/PIM/Dialect/Spatial/SpatialOps.hpp"
 #include "src/Dialect/ONNX/ONNXOps.hpp"
 
@@ -49,6 +50,45 @@ materializeScaledConstantTensor(Value value, float factor, ConversionPatternRewr
   return arith::ConstantOp::create(rewriter, loc, denseAttr.getType(), scaledAttr).getResult();
 }
 
+static Value transposeForSpatial(Value value,
+                                 RankedTensorType resultType,
+                                 ArrayRef<int64_t> permutation,
+                                 ConversionPatternRewriter& rewriter,
+                                 Location loc) {
+  if (isHostFoldableValue(value))
+    return ONNXTransposeOp::create(rewriter, loc, resultType, value, rewriter.getI64ArrayAttr(permutation));
+
+  auto computeOp = createSpatCompute<1>(rewriter, loc, TypeRange {resultType}, {}, value, [&](Value input) {
+    Value transposed = ONNXTransposeOp::create(rewriter, loc, resultType, input, rewriter.getI64ArrayAttr(permutation));
+    spatial::SpatYieldOp::create(rewriter, loc, transposed);
+  });
+  return computeOp.getResult(0);
+}
+
+static Value
+expandRankOneBias(Value value, RankedTensorType resultType, ConversionPatternRewriter& rewriter, Location loc) {
+  if (isHostFoldableValue(value))
+    return tensor::ExpandShapeOp::create(rewriter,
+                                         loc,
+                                         resultType,
+                                         value,
+                                         SmallVector<ReassociationIndices> {
+                                           {0, 1}
+    });
+
+  auto computeOp = createSpatCompute<1>(rewriter, loc, TypeRange {resultType}, {}, value, [&](Value input) {
+    Value expanded = tensor::ExpandShapeOp::create(rewriter,
+                                                   loc,
+                                                   resultType,
+                                                   input,
+                                                   SmallVector<ReassociationIndices> {
+                                                     {0, 1}
+    });
+    spatial::SpatYieldOp::create(rewriter, loc, expanded);
+  });
+  return computeOp.getResult(0);
+}
+
 struct GemmToManyGemv : OpConversionPattern<ONNXGemmOp> {
   using OpConversionPattern::OpConversionPattern;
 
@@ -65,6 +105,72 @@ struct GemvToSpatialCompute : OpConversionPattern<ONNXGemmOp> {
                                 ConversionPatternRewriter& rewriter) const override;
 };
 
+struct GemmToSpatialComputeBatch : OpConversionPattern<ONNXGemmOp> {
+  using OpConversionPattern::OpConversionPattern;
+
+  LogicalResult matchAndRewrite(ONNXGemmOp gemmOp,
+                                ONNXGemmOpAdaptor gemmOpAdaptor,
+                                ConversionPatternRewriter& rewriter) const override;
+};
+
+static SmallVector<Value> materializeBatchRowSlices(Value matrix,
+                                                    RankedTensorType matrixType,
+                                                    ConversionPatternRewriter& rewriter,
+                                                    Location loc) {
+  const int64_t numRows = matrixType.getDimSize(0);
+  auto rowType = RankedTensorType::get({1, matrixType.getDimSize(1)}, matrixType.getElementType());
+  SmallVector<Type> resultTypes(static_cast<size_t>(numRows), rowType);
+
+  if (isHostFoldableValue(matrix)) {
+    auto extractRowsOp = spatial::SpatExtractRowsOp::create(rewriter, loc, TypeRange(resultTypes), matrix);
+    return SmallVector<Value>(extractRowsOp->result_begin(), extractRowsOp->result_end());
+  }
+
+  auto buildRowSlices = [&](Value matrixArg) {
+    auto extractRowsOp = spatial::SpatExtractRowsOp::create(rewriter, loc, TypeRange(resultTypes), matrixArg);
+    return SmallVector<Value>(extractRowsOp->result_begin(), extractRowsOp->result_end());
+  };
+
+  auto cloneBatchInputChainIntoSliceCompute =
+    [&](Value rootInput, SmallVector<Operation*> chainOps, Value rootValue) -> SmallVector<Value> {
+    auto sliceCompute =
+      createSpatCompute<1>(rewriter, loc, TypeRange(resultTypes), {}, ValueRange {rootInput}, [&](Value input) {
+        Value transformedMatrix = input;
+        if (!chainOps.empty()) {
+          IRMapping mapper;
+          mapper.map(rootValue, input);
+          for (Operation* chainOp : chainOps)
+            rewriter.clone(*chainOp, mapper);
+          transformedMatrix = cast<Value>(mapper.lookup(matrix));
+        }
+        spatial::SpatYieldOp::create(rewriter, loc, buildRowSlices(transformedMatrix));
+      });
+    SmallVector<Value> rowSlices(sliceCompute->result_begin(), sliceCompute->result_end());
+    return rowSlices;
+  };
+
+  SmallVector<Operation*> chainOps;
+  Value rootValue = matrix;
+  while (Operation* definingOp = rootValue.getDefiningOp()) {
+    if (auto rootCompute = dyn_cast<spatial::SpatCompute>(definingOp)) {
+      SmallVector<Operation*> reversedChainOps(chainOps.rbegin(), chainOps.rend());
+      return cloneBatchInputChainIntoSliceCompute(
+        rootCompute.getResult(cast<OpResult>(rootValue).getResultNumber()), reversedChainOps, rootValue);
+    }
+
+    if (definingOp->getNumOperands() != 1)
+      break;
+    if (!isa<tensor::ExtractSliceOp, tensor::ExpandShapeOp, tensor::CollapseShapeOp, ONNXTransposeOp>(definingOp))
+      break;
+
+    chainOps.push_back(definingOp);
+    rootValue = definingOp->getOperand(0);
+  }
+
+  SmallVector<Operation*> reversedChainOps(chainOps.rbegin(), chainOps.rend());
+  return cloneBatchInputChainIntoSliceCompute(rootValue, reversedChainOps, rootValue);
+}
+
 } // namespace
 
 LogicalResult GemmToManyGemv::matchAndRewrite(ONNXGemmOp gemmOp,
@@ -75,13 +181,23 @@ LogicalResult GemmToManyGemv::matchAndRewrite(ONNXGemmOp gemmOp,
   Value b = gemmOpAdaptor.getB();
   Value c = gemmOpAdaptor.getC();
 
-  assert("A should have been transposed already" && !gemmOpAdaptor.getTransA());
+  if (gemmOpAdaptor.getTransA()) {
+    gemmOp.emitOpError("requires transA=false before Gemm row decomposition");
+    return failure();
+  }
 
   bool hasC = !isa<ONNXNoneOp>(c.getDefiningOp());
 
   auto aType = cast<RankedTensorType>(a.getType());
   auto outType = cast<RankedTensorType>(gemmOp.getY().getType());
-  assert("Only support static shapes" && aType.hasStaticShape() && outType.hasStaticShape());
+  if (!aType.hasStaticShape()) {
+    pim::emitUnsupportedStaticShapeDiagnostic(gemmOp, "Gemm input A");
+    return failure();
+  }
+  if (!outType.hasStaticShape()) {
+    pim::emitUnsupportedStaticShapeDiagnostic(gemmOp, "Gemm result");
+    return failure();
+  }
 
   const int64_t numOutRows = aType.getDimSize(0);
 
@@ -105,47 +221,43 @@ LogicalResult GemmToManyGemv::matchAndRewrite(ONNXGemmOp gemmOp,
     // Expand rank-1 bias [N] to rank-2 [1, N] for uniform handling
     if (cType.getRank() == 1) {
       auto expandedType = RankedTensorType::get({1, cType.getDimSize(0)}, cType.getElementType());
-      c = tensor::ExpandShapeOp::create(rewriter,
+      c = expandRankOneBias(c, expandedType, rewriter, loc);
-                                        loc,
-                                        expandedType,
-                                        c,
-                                        SmallVector<ReassociationIndices> {
-                                          {0, 1}
-      });
       cType = expandedType;
     }
-    assert("Only support rank 2 tensor for C" && cType.getRank() == 2);
+    if (!cType.hasStaticShape()) {
+      pim::emitUnsupportedStaticShapeDiagnostic(gemmOp, "Gemm bias");
+      return failure();
+    }
+    if (cType.getRank() != 2) {
+      pim::emitUnsupportedRankDiagnostic(gemmOp, "Gemm bias", cType.getRank(), {1, 2});
+      return failure();
+    }
     cHasNumOutRows = cType.getDimSize(0) == numOutRows;
   }
 
   auto outRowType = RankedTensorType::get({1, outType.getDimSize(1)}, outType.getElementType());
+  SmallVector<Value> aSlices = materializeBatchRowSlices(a, aType, rewriter, loc);
+  SmallVector<Value> cSlices;
+  if (hasC && cHasNumOutRows)
+    cSlices = materializeBatchRowSlices(c, cType, rewriter, loc);
 
   SmallVector<Value> gemvOps;
-  gemvOps.reserve(numOutRows);
+  gemvOps.reserve(static_cast<size_t>(numOutRows));
   for (int64_t rowIdx = 0; rowIdx < numOutRows; rowIdx++) {
-    SmallVector<OpFoldResult> offsets = {rewriter.getIndexAttr(rowIdx), rewriter.getIndexAttr(0)};
-    SmallVector<OpFoldResult> sizes = {rewriter.getIndexAttr(1), rewriter.getIndexAttr(aType.getDimSize(1))};
-    SmallVector<OpFoldResult> strides = {rewriter.getIndexAttr(1), rewriter.getIndexAttr(1)};
-    auto aSliceType = RankedTensorType::get({1, aType.getDimSize(1)}, aType.getElementType());
-    auto aSlice = tensor::ExtractSliceOp::create(rewriter, loc, aSliceType, a, offsets, sizes, strides).getResult();
-
     Value cSlice = c;
     if (hasC) {
-      if (cHasNumOutRows) {
+      if (cHasNumOutRows)
-        SmallVector<OpFoldResult> offsets = {rewriter.getIndexAttr(rowIdx), rewriter.getIndexAttr(0)};
+        cSlice = cSlices[static_cast<size_t>(rowIdx)];
-        SmallVector<OpFoldResult> sizes = {rewriter.getIndexAttr(1), rewriter.getIndexAttr(cType.getDimSize(1))};
+      else if (!isVectorShape(getTensorShape(c))) {
-        SmallVector<OpFoldResult> strides = {rewriter.getIndexAttr(1), rewriter.getIndexAttr(1)};
+        gemmOp.emitOpError("requires Gemm bias C to be vector-like when shared across decomposed rows");
-        auto cSliceType = RankedTensorType::get({1, cType.getDimSize(1)}, cType.getElementType());
+        return failure();
-        cSlice = tensor::ExtractSliceOp::create(rewriter, loc, cSliceType, c, offsets, sizes, strides).getResult();
       }
-      else
-        assert("C should be a vector" && isVectorShape(getTensorShape(c)));
     }
 
     auto gemvOp = ONNXGemmOp::create(rewriter,
                                      loc,
                                      outRowType,
-                                     aSlice,
+                                     aSlices[static_cast<size_t>(rowIdx)],
                                      b,
                                      cSlice,
                                      rewriter.getF32FloatAttr(1.0f),
@@ -156,8 +268,7 @@ LogicalResult GemmToManyGemv::matchAndRewrite(ONNXGemmOp gemmOp,
   }
 
   auto concatComputeOp = createSpatCompute(rewriter, loc, gemmOp.getType(), {}, gemvOps, [&](ValueRange gemvOpsArgs) {
-    auto concatOp = tensor::ConcatOp::create(rewriter, loc, /*axis=*/0, gemvOpsArgs);
+    spatial::SpatYieldOp::create(rewriter, loc, createSpatConcat(rewriter, loc, /*axis=*/0, gemvOpsArgs));
-    spatial::SpatYieldOp::create(rewriter, loc, concatOp.getResult());
   });
 
   rewriter.replaceOp(gemmOp, concatComputeOp);
@@ -189,20 +300,31 @@ LogicalResult GemvToSpatialCompute::matchAndRewrite(ONNXGemmOp gemmOp,
     // Expand rank-1 bias [N] to rank-2 [1, N] for uniform handling
     if (cType.getRank() == 1) {
       auto expandedType = RankedTensorType::get({1, cType.getDimSize(0)}, cType.getElementType());
-      c = tensor::ExpandShapeOp::create(rewriter,
+      c = expandRankOneBias(c, expandedType, rewriter, gemmLoc);
-                                        gemmLoc,
-                                        expandedType,
-                                        c,
-                                        SmallVector<ReassociationIndices> {
-                                          {0, 1}
-      });
       cType = expandedType;
     }
-    assert("Only support rank 2 tensor for C" && cType.getRank() == 2);
+    if (!cType.hasStaticShape()) {
+      pim::emitUnsupportedStaticShapeDiagnostic(gemmOp, "Gemm bias");
+      return failure();
+    }
+    if (cType.getRank() != 2) {
+      pim::emitUnsupportedRankDiagnostic(gemmOp, "Gemm bias", cType.getRank(), {1, 2});
+      return failure();
+    }
   }
 
-  assert("Only support static shapes" && aType.hasStaticShape() && bType.hasStaticShape()
+  if (!aType.hasStaticShape()) {
-         && (!hasC || cType.hasStaticShape()) && outType.hasStaticShape());
+    pim::emitUnsupportedStaticShapeDiagnostic(gemmOp, "Gemm input A");
+    return failure();
+  }
+  if (!bType.hasStaticShape()) {
+    pim::emitUnsupportedStaticShapeDiagnostic(gemmOp, "Gemm input B");
+    return failure();
+  }
+  if (!outType.hasStaticShape()) {
+    pim::emitUnsupportedStaticShapeDiagnostic(gemmOp, "Gemm result");
+    return failure();
+  }
 
   if (!isVectorShape(aType.getShape()) || (hasC && !isVectorShape(cType.getShape())))
     // Not a gemv
@@ -210,13 +332,14 @@ LogicalResult GemvToSpatialCompute::matchAndRewrite(ONNXGemmOp gemmOp,
 
   if (transA) {
     auto aShape = aType.getShape();
-    auto transposedType = aType.cloneWith(ArrayRef({aShape[1], aShape[0]}), aType.getElementType());
+    auto transposedType = RankedTensorType::get({aShape[1], aShape[0]}, aType.getElementType());
-    a = ONNXTransposeOp::create(rewriter, gemmLoc, transposedType, a, rewriter.getI64ArrayAttr({1, 0}));
+    a = transposeForSpatial(a, transposedType, {1, 0}, rewriter, gemmLoc);
+    aType = cast<RankedTensorType>(a.getType());
   }
   if (transB) {
     auto bShape = bType.getShape();
-    auto transposedType = bType.cloneWith(ArrayRef({bShape[1], bShape[0]}), bType.getElementType());
+    auto transposedType = RankedTensorType::get({bShape[1], bShape[0]}, bType.getElementType());
-    b = ONNXTransposeOp::create(rewriter, gemmLoc, transposedType, b, rewriter.getI64ArrayAttr({1, 0}));
+    b = transposeForSpatial(b, transposedType, {1, 0}, rewriter, gemmLoc);
     bType = cast<RankedTensorType>(b.getType());
   }
 
@@ -240,7 +363,6 @@ LogicalResult GemvToSpatialCompute::matchAndRewrite(ONNXGemmOp gemmOp,
   auto [aNumHSlices, aLastHSliceSize] = ceilIntegerDivideWithRemainder(aType.getDimSize(1), crossbarSize.getValue());
   auto [bNumHSlices, bLastHSliceSize] = ceilIntegerDivideWithRemainder(bType.getDimSize(1), crossbarSize.getValue());
   auto bNumVSlices = aNumHSlices;
-  auto bLastVSliceSize = aLastHSliceSize;
   auto cNumHSlices = bNumHSlices;
   auto cLastHSliceSize = bLastHSliceSize;
   auto outNumHSlices = cNumHSlices;
@@ -281,19 +403,25 @@ LogicalResult GemvToSpatialCompute::matchAndRewrite(ONNXGemmOp gemmOp,
         weights.push_back(bTiles[outSliceId][coreId][aSliceId]);
 
       auto computeOp = createSpatCompute(
-        rewriter, gemmLoc, currOutHSliceType, weights, aHSlices[coreId], [&](ValueRange aHSlicesArgs) {
+        rewriter, gemmLoc, currOutHSliceType, weights, aHSlices[coreId], [&](ValueRange aHSlicesArgs) -> LogicalResult {
           SmallVector<Value> vmmOutputs;
           vmmOutputs.reserve(aHSlicesArgs.size());
           for (auto [aHSliceId, computeArg] : llvm::enumerate(aHSlicesArgs))
             vmmOutputs.push_back(
-              spatial::SpatWeightedVMMOp::create(rewriter, gemmLoc, currOutHSliceType, aHSliceId, computeArg));
+              spatial::SpatVMMOp::create(rewriter, gemmLoc, currOutHSliceType, aHSliceId, computeArg));
-          assert(!vmmOutputs.empty() && "vmmOutputs must be non-empty");
+          if (vmmOutputs.empty()) {
+            gemmOp.emitOpError("requires at least one non-empty slice when lowering tiled Gemm to Spatial VMMs");
+            return failure();
+          }
 
           Value partialVmmSum = sumTensors(vmmOutputs, rewriter);
           spatial::SpatYieldOp::create(rewriter, gemmLoc, partialVmmSum);
+          return success();
         });
+      if (failed(computeOp))
+        return failure();
 
-      partialResults.push_back(computeOp.getResult(0));
+      partialResults.push_back(computeOp->getResult(0));
     }
 
     if (hasC) {
@@ -313,8 +441,126 @@ LogicalResult GemvToSpatialCompute::matchAndRewrite(ONNXGemmOp gemmOp,
 
   auto concatComputeOp =
     createSpatCompute(rewriter, gemmLoc, gemmOp.getType(), {}, outHSlices, [&](ValueRange blockArgs) {
-      auto concatOp = tensor::ConcatOp::create(rewriter, gemmLoc, /*axis=*/1, blockArgs);
+      spatial::SpatYieldOp::create(rewriter, gemmLoc, createSpatConcat(rewriter, gemmLoc, /*axis=*/1, blockArgs));
-      spatial::SpatYieldOp::create(rewriter, gemmLoc, concatOp.getResult());
+    });
+
+  rewriter.replaceOp(gemmOp, concatComputeOp);
+  return success();
+}
+
+LogicalResult GemmToSpatialComputeBatch::matchAndRewrite(ONNXGemmOp gemmOp,
+                                                         ONNXGemmOpAdaptor gemmOpAdaptor,
+                                                         ConversionPatternRewriter& rewriter) const {
+  Location loc = gemmOp.getLoc();
+  Value a = gemmOpAdaptor.getA();
+  Value b = gemmOpAdaptor.getB();
+  Value c = gemmOpAdaptor.getC();
+
+  if (gemmOpAdaptor.getTransA()) {
+    gemmOp.emitOpError("requires transA=false before batch Gemm lowering");
+    return failure();
+  }
+
+  bool hasC = !isa<ONNXNoneOp>(c.getDefiningOp());
+
+  auto aType = cast<RankedTensorType>(a.getType());
+  auto bType = cast<RankedTensorType>(b.getType());
+  auto outType = cast<RankedTensorType>(gemmOp.getY().getType());
+  if (!aType.hasStaticShape()) {
+    pim::emitUnsupportedStaticShapeDiagnostic(gemmOp, "Gemm input A");
+    return failure();
+  }
+  if (!bType.hasStaticShape()) {
+    pim::emitUnsupportedStaticShapeDiagnostic(gemmOp, "Gemm input B");
+    return failure();
+  }
+  if (!outType.hasStaticShape()) {
+    pim::emitUnsupportedStaticShapeDiagnostic(gemmOp, "Gemm result");
+    return failure();
+  }
+
+  const int64_t numOutRows = aType.getDimSize(0);
+  if (numOutRows <= 1)
+    return failure();
+
+  // Only handle the single-tile case: K <= crossbarSize and N <= crossbarSize
+  if (aType.getDimSize(1) > static_cast<int64_t>(crossbarSize.getValue())
+      || outType.getDimSize(1) > static_cast<int64_t>(crossbarSize.getValue()))
+    return failure();
+
+  auto scaledB = materializeScaledConstantTensor(b, gemmOpAdaptor.getAlpha().convertToFloat(), rewriter, loc);
+  if (failed(scaledB))
+    return failure();
+  b = *scaledB;
+  bType = cast<RankedTensorType>(b.getType());
+
+  if (gemmOpAdaptor.getTransB()) {
+    auto bShape = bType.getShape();
+    auto transposedType = RankedTensorType::get({bShape[1], bShape[0]}, bType.getElementType());
+    b = transposeForSpatial(b, transposedType, {1, 0}, rewriter, loc);
+    bType = cast<RankedTensorType>(b.getType());
+  }
+  (void) bType;
+
+  if (!isHostFoldableValue(b))
+    return failure();
+
+  Value sharedBias;
+  if (hasC) {
+    auto scaledC = materializeScaledConstantTensor(c, gemmOpAdaptor.getBeta().convertToFloat(), rewriter, loc);
+    if (failed(scaledC))
+      return failure();
+    c = *scaledC;
+    auto cType = cast<RankedTensorType>(c.getType());
+    if (cType.getRank() == 1) {
+      auto expandedType = RankedTensorType::get({1, cType.getDimSize(0)}, cType.getElementType());
+      c = expandRankOneBias(c, expandedType, rewriter, loc);
+      cType = cast<RankedTensorType>(c.getType());
+    }
+    if (!cType.hasStaticShape()) {
+      pim::emitUnsupportedStaticShapeDiagnostic(gemmOp, "Gemm bias");
+      return failure();
+    }
+    if (cType.getRank() != 2) {
+      pim::emitUnsupportedRankDiagnostic(gemmOp, "Gemm bias", cType.getRank(), {1, 2});
+      return failure();
+    }
+    // Row-specific bias can't share a single template body; fall through to GemmToManyGemv
+    if (cType.getDimSize(0) == numOutRows && numOutRows > 1)
+      return failure();
+    if (cType.getDimSize(0) == 1 && cType.getDimSize(1) == 1)
+      c = broadcastToVector(c, outType.getDimSize(1), rewriter, loc);
+    sharedBias = c;
+  }
+
+  SmallVector<Value> aSlices = materializeBatchRowSlices(a, aType, rewriter, loc);
+  auto aSliceType = cast<RankedTensorType>(aSlices.front().getType());
+
+  auto outRowType = RankedTensorType::get({1, outType.getDimSize(1)}, outType.getElementType());
+  SmallVector<Type> resultTypes(static_cast<size_t>(numOutRows), outRowType);
+  SmallVector<Value> weights(static_cast<size_t>(numOutRows), b);
+
+  auto batchOp = spatial::SpatComputeBatch::create(rewriter,
+                                                   loc,
+                                                   TypeRange(resultTypes),
+                                                   rewriter.getI32IntegerAttr(static_cast<int32_t>(numOutRows)),
+                                                   ValueRange(weights),
+                                                   ValueRange(aSlices));
+
+  Block* body = rewriter.createBlock(
+    &batchOp.getBody(), batchOp.getBody().end(), TypeRange {aSliceType}, SmallVector<Location>(1, loc));
+  rewriter.setInsertionPointToEnd(body);
+
+  Value vmmResult = spatial::SpatVMMOp::create(rewriter, loc, outRowType, 0, body->getArgument(0)).getResult();
+  Value laneResult = vmmResult;
+  if (sharedBias)
+    laneResult = spatial::SpatVAddOp::create(rewriter, loc, outRowType, vmmResult, sharedBias).getResult();
+  spatial::SpatYieldOp::create(rewriter, loc, laneResult);
+
+  rewriter.setInsertionPointAfter(batchOp);
+  SmallVector<Value> laneResults(batchOp->result_begin(), batchOp->result_end());
+  auto concatComputeOp = createSpatCompute(rewriter, loc, gemmOp.getType(), {}, laneResults, [&](ValueRange args) {
+    spatial::SpatYieldOp::create(rewriter, loc, createSpatConcat(rewriter, loc, /*axis=*/0, args));
   });
 
   rewriter.replaceOp(gemmOp, concatComputeOp);
@@ -322,6 +568,7 @@ LogicalResult GemvToSpatialCompute::matchAndRewrite(ONNXGemmOp gemmOp,
 }
 
 void populateGemmPatterns(RewritePatternSet& patterns, MLIRContext* ctx) {
+  patterns.insert<GemmToSpatialComputeBatch>(ctx, PatternBenefit(2));
   patterns.insert<GemmToManyGemv>(ctx);
   patterns.insert<GemvToSpatialCompute>(ctx);
 }

src/PIM/Conversion/ONNXToSpatial/Patterns/Math/MatMul.cpp

@@ -4,8 +4,9 @@
 
 #include "llvm/ADT/SmallVector.h"
 
-#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Common.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Common/Common.hpp"
-#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Patterns.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/ConversionPatterns.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/HostFoldability.hpp"
 #include "src/Accelerators/PIM/Dialect/Spatial/SpatialOps.hpp"
 #include "src/Dialect/ONNX/ONNXOps.hpp"
 
@@ -14,7 +15,102 @@ using namespace mlir;
 namespace onnx_mlir {
 namespace {
 
-struct MatMulRank3ToGemm : OpRewritePattern<ONNXMatMulOp> {
+static bool haveStaticPositiveShape(ArrayRef<int64_t> shape) {
+  return llvm::all_of(shape, [](int64_t dim) { return dim > 0; });
+}
+
+static Value extractBatchMatrix(Value value,
+                                int64_t batchIndex,
+                                int64_t batchSize,
+                                int64_t rows,
+                                int64_t cols,
+                                PatternRewriter& rewriter,
+                                Location loc) {
+  auto type = cast<RankedTensorType>(value.getType());
+  if (type.getRank() == 2)
+    return value;
+
+  auto sliceType = RankedTensorType::get({1, rows, cols}, type.getElementType());
+  SmallVector<OpFoldResult> offsets = {
+    rewriter.getIndexAttr(batchSize == 1 ? 0 : batchIndex), rewriter.getIndexAttr(0), rewriter.getIndexAttr(0)};
+  SmallVector<OpFoldResult> sizes = {
+    rewriter.getIndexAttr(1), rewriter.getIndexAttr(rows), rewriter.getIndexAttr(cols)};
+  SmallVector<OpFoldResult> strides = {rewriter.getIndexAttr(1), rewriter.getIndexAttr(1), rewriter.getIndexAttr(1)};
+  auto matrixType = RankedTensorType::get({rows, cols}, type.getElementType());
+  auto buildMatrix = [&](Value input) -> Value {
+    Value slice = tensor::ExtractSliceOp::create(rewriter, loc, sliceType, input, offsets, sizes, strides);
+    return tensor::CollapseShapeOp::create(rewriter,
+                                           loc,
+                                           matrixType,
+                                           slice,
+                                           SmallVector<ReassociationIndices> {
+                                             {0, 1},
+                                             {2}
+    });
+  };
+
+  if (isHostFoldableValue(value))
+    return buildMatrix(value);
+
+  auto batchMatrixCompute =
+    createSpatCompute<1>(rewriter, loc, TypeRange {matrixType}, {}, ValueRange {value}, [&](Value input) {
+      spatial::SpatYieldOp::create(rewriter, loc, buildMatrix(input));
+    });
+  return batchMatrixCompute.getResult(0);
+}
+
+static Value transposeLastTwoDims(Value value, PatternRewriter& rewriter, Location loc) {
+  auto type = cast<RankedTensorType>(value.getType());
+  auto shape = type.getShape();
+  if (type.getRank() == 2) {
+    auto transposedType = RankedTensorType::get({shape[1], shape[0]}, type.getElementType());
+    return ONNXTransposeOp::create(rewriter, loc, transposedType, value, rewriter.getI64ArrayAttr({1, 0}));
+  }
+
+  auto transposedType = RankedTensorType::get({shape[0], shape[2], shape[1]}, type.getElementType());
+  return ONNXTransposeOp::create(rewriter, loc, transposedType, value, rewriter.getI64ArrayAttr({0, 2, 1}));
+}
+
+static Value transposeLastTwoDimsInCompute(Value value, PatternRewriter& rewriter, Location loc) {
+  auto type = cast<RankedTensorType>(value.getType());
+  auto shape = type.getShape();
+  RankedTensorType transposedType;
+  SmallVector<int64_t> perm;
+  if (type.getRank() == 2) {
+    transposedType = RankedTensorType::get({shape[1], shape[0]}, type.getElementType());
+    perm = {1, 0};
+  }
+  else {
+    transposedType = RankedTensorType::get({shape[0], shape[2], shape[1]}, type.getElementType());
+    perm = {0, 2, 1};
+  }
+
+  auto transposeCompute = createSpatCompute<1>(rewriter, loc, transposedType, {}, ValueRange {value}, [&](Value input) {
+    Value transposed = ONNXTransposeOp::create(rewriter, loc, transposedType, input, rewriter.getI64ArrayAttr(perm));
+    spatial::SpatYieldOp::create(rewriter, loc, transposed);
+  });
+  return transposeCompute.getResult(0);
+}
+
+static Value concatValues(ValueRange inputs, int64_t axis, PatternRewriter& rewriter, Location loc) {
+  auto firstType = cast<RankedTensorType>(inputs.front().getType());
+  SmallVector<int64_t> outputShape(firstType.getShape().begin(), firstType.getShape().end());
+  int64_t concatDimSize = 0;
+  for (Value input : inputs)
+    concatDimSize += cast<RankedTensorType>(input.getType()).getDimSize(axis);
+  outputShape[axis] = concatDimSize;
+  auto resultType = RankedTensorType::get(outputShape, firstType.getElementType(), firstType.getEncoding());
+
+  if (llvm::all_of(inputs, isHostFoldableValue))
+    return createSpatConcat(rewriter, loc, axis, inputs);
+
+  auto concatCompute = createSpatCompute(rewriter, loc, TypeRange {resultType}, {}, inputs, [&](ValueRange args) {
+    spatial::SpatYieldOp::create(rewriter, loc, createSpatConcat(rewriter, loc, axis, args));
+  });
+  return concatCompute.getResult(0);
+}
+
+struct MatMulToGemm : OpRewritePattern<ONNXMatMulOp> {
   using OpRewritePattern::OpRewritePattern;
 
   LogicalResult matchAndRewrite(ONNXMatMulOp matmulOp, PatternRewriter& rewriter) const override {
@@ -24,80 +120,125 @@ struct MatMulRank3ToGemm : OpRewritePattern<ONNXMatMulOp> {
     if (!lhsType || !rhsType || !outType || !lhsType.hasStaticShape() || !rhsType.hasStaticShape()
         || !outType.hasStaticShape())
       return failure();
-    if (lhsType.getRank() != 2 || rhsType.getRank() != 3 || outType.getRank() != 3)
+    if ((lhsType.getRank() != 2 && lhsType.getRank() != 3) || (rhsType.getRank() != 2 && rhsType.getRank() != 3)
+        || (outType.getRank() != 2 && outType.getRank() != 3))
+      return failure();
+    if (!haveStaticPositiveShape(lhsType.getShape()) || !haveStaticPositiveShape(rhsType.getShape())
+        || !haveStaticPositiveShape(outType.getShape()))
       return failure();
 
-    const int64_t batch = rhsType.getDimSize(0);
+    const int64_t lhsBatch = lhsType.getRank() == 3 ? lhsType.getDimSize(0) : 1;
-    const int64_t k = rhsType.getDimSize(1);
+    const int64_t rhsBatch = rhsType.getRank() == 3 ? rhsType.getDimSize(0) : 1;
-    const int64_t n = rhsType.getDimSize(2);
+    const int64_t batch = std::max(lhsBatch, rhsBatch);
-    const int64_t m = lhsType.getDimSize(0);
+
-    if (lhsType.getDimSize(1) != k || outType.getDimSize(0) != batch || outType.getDimSize(1) != m
+    if ((lhsBatch != 1 && lhsBatch != batch) || (rhsBatch != 1 && rhsBatch != batch))
-        || outType.getDimSize(2) != n)
       return failure();
 
+    const int64_t m = lhsType.getRank() == 3 ? lhsType.getDimSize(1) : lhsType.getDimSize(0);
+    const int64_t k = lhsType.getRank() == 3 ? lhsType.getDimSize(2) : lhsType.getDimSize(1);
+    const int64_t rhsK = rhsType.getRank() == 3 ? rhsType.getDimSize(1) : rhsType.getDimSize(0);
+    const int64_t n = rhsType.getRank() == 3 ? rhsType.getDimSize(2) : rhsType.getDimSize(1);
+    if (k != rhsK)
+      return failure();
+
+    if (outType.getRank() == 2) {
+      if (batch != 1 || outType.getDimSize(0) != m || outType.getDimSize(1) != n)
+        return failure();
+    }
+    else {
+      if (outType.getDimSize(0) != batch || outType.getDimSize(1) != m || outType.getDimSize(2) != n)
+        return failure();
+    }
+
     Location loc = matmulOp.getLoc();
-    auto lhsTransposedType = RankedTensorType::get({k, m}, lhsType.getElementType());
+    bool useTransposedForm = isHostFoldableValue(matmulOp.getA()) && !isHostFoldableValue(matmulOp.getB());
-    auto rhsSliceType = RankedTensorType::get({1, k, 1}, rhsType.getElementType());
-    auto rhsRowType = RankedTensorType::get({1, k}, rhsType.getElementType());
-    auto gemmRowType = RankedTensorType::get({1, m}, outType.getElementType());
-    auto gemmOutType = RankedTensorType::get({batch * n, m}, outType.getElementType());
-    auto gemmExpandedType = RankedTensorType::get({batch, n, m}, outType.getElementType());
 
-    Value lhsTransposed =
+    Value lhs = matmulOp.getA();
-      ONNXTransposeOp::create(rewriter, loc, lhsTransposedType, matmulOp.getA(), rewriter.getI64ArrayAttr({1, 0}));
+    Value rhs = matmulOp.getB();
+    int64_t lhsBatchForGemm = lhsBatch;
+    int64_t rhsBatchForGemm = rhsBatch;
+    int64_t gemmM = m;
+    int64_t gemmK = k;
+    int64_t gemmN = n;
+    if (useTransposedForm) {
+      lhs = transposeLastTwoDimsInCompute(matmulOp.getB(), rewriter, loc);
+      lhsBatchForGemm = rhsBatch;
+      rhs = transposeLastTwoDims(matmulOp.getA(), rewriter, loc);
+      rhsBatchForGemm = lhsBatch;
+      gemmM = n;
+      gemmN = m;
+    }
+
+    auto gemmType = RankedTensorType::get({gemmM, gemmN}, outType.getElementType());
+    auto batchedOutType = RankedTensorType::get({1, m, n}, outType.getElementType());
     Value none = ONNXNoneOp::create(rewriter, loc, rewriter.getNoneType());
 
-    SmallVector<Value> gemmRows;
+    if (outType.getRank() == 2) {
-    gemmRows.reserve(batch * n);
+      Value lhsMatrix = extractBatchMatrix(lhs, /*batchIndex=*/0, lhsBatchForGemm, gemmM, gemmK, rewriter, loc);
-    for (int64_t batchIdx = 0; batchIdx < batch; batchIdx++) {
+      Value rhsMatrix = extractBatchMatrix(rhs, /*batchIndex=*/0, rhsBatchForGemm, gemmK, gemmN, rewriter, loc);
-      for (int64_t colIdx = 0; colIdx < n; colIdx++) {
+      Value gemmResult = ONNXGemmOp::create(rewriter,
-        SmallVector<OpFoldResult> offsets = {
-          rewriter.getIndexAttr(batchIdx), rewriter.getIndexAttr(0), rewriter.getIndexAttr(colIdx)};
-        SmallVector<OpFoldResult> sizes = {
-          rewriter.getIndexAttr(1), rewriter.getIndexAttr(k), rewriter.getIndexAttr(1)};
-        SmallVector<OpFoldResult> strides = {
-          rewriter.getIndexAttr(1), rewriter.getIndexAttr(1), rewriter.getIndexAttr(1)};
-        Value rhsSlice =
-          tensor::ExtractSliceOp::create(rewriter, loc, rhsSliceType, matmulOp.getB(), offsets, sizes, strides);
-        Value rhsRow = tensor::CollapseShapeOp::create(rewriter,
                                             loc,
-                                                       rhsRowType,
+                                            gemmType,
-                                                       rhsSlice,
+                                            lhsMatrix,
-                                                       SmallVector<ReassociationIndices> {
+                                            rhsMatrix,
-                                                         {0},
-                                                         {1, 2}
-        });
-
-        auto gemmOp = ONNXGemmOp::create(rewriter,
-                                         loc,
-                                         gemmRowType,
-                                         rhsRow,
-                                         lhsTransposed,
                                             none,
                                             rewriter.getF32FloatAttr(1.0f),
                                             rewriter.getF32FloatAttr(1.0f),
                                             rewriter.getBoolAttr(false),
-                                         rewriter.getBoolAttr(false));
+                                            rewriter.getBoolAttr(false))
-        gemmRows.push_back(gemmOp.getY());
+                           .getY();
-      }
+      if (useTransposedForm) {
-    }
+        auto transposeCompute =
-
+          createSpatCompute<1>(rewriter, loc, TypeRange {outType}, {}, gemmResult, [&](Value input) {
-    auto concatComputeOp = createSpatCompute(rewriter, loc, gemmOutType, {}, gemmRows, [&](ValueRange gemmRowsArgs) {
+            Value transposed = ONNXTransposeOp::create(rewriter, loc, outType, input, rewriter.getI64ArrayAttr({1, 0}));
-      auto concatOp = tensor::ConcatOp::create(rewriter, loc, /*axis=*/0, gemmRowsArgs);
+            spatial::SpatYieldOp::create(rewriter, loc, transposed);
-      spatial::SpatYieldOp::create(rewriter, loc, concatOp.getResult());
           });
+        gemmResult = transposeCompute.getResult(0);
+      }
+      rewriter.replaceOp(matmulOp, gemmResult);
+      return success();
+    }
 
-    Value gemmOut = concatComputeOp.getResult(0);
+    SmallVector<Value> batchResults;
-    Value gemmExpanded = tensor::ExpandShapeOp::create(rewriter,
+    batchResults.reserve(batch);
+    for (int64_t batchIdx = 0; batchIdx < batch; batchIdx++) {
+      Value lhsMatrix = extractBatchMatrix(lhs, batchIdx, lhsBatchForGemm, gemmM, gemmK, rewriter, loc);
+      Value rhsMatrix = extractBatchMatrix(rhs, batchIdx, rhsBatchForGemm, gemmK, gemmN, rewriter, loc);
+      Value gemmResult = ONNXGemmOp::create(rewriter,
                                             loc,
-                                                       gemmExpandedType,
+                                            gemmType,
-                                                       gemmOut,
+                                            lhsMatrix,
+                                            rhsMatrix,
+                                            none,
+                                            rewriter.getF32FloatAttr(1.0f),
+                                            rewriter.getF32FloatAttr(1.0f),
+                                            rewriter.getBoolAttr(false),
+                                            rewriter.getBoolAttr(false))
+                           .getY();
+      auto batchResultCompute =
+        createSpatCompute<1>(rewriter, loc, TypeRange {batchedOutType}, {}, gemmResult, [&](Value input) {
+          Value resultMatrix = input;
+          if (useTransposedForm) {
+            resultMatrix = ONNXTransposeOp::create(rewriter,
+                                                   loc,
+                                                   RankedTensorType::get({m, n}, outType.getElementType()),
+                                                   input,
+                                                   rewriter.getI64ArrayAttr({1, 0}));
+          }
+          Value expanded = tensor::ExpandShapeOp::create(rewriter,
+                                                         loc,
+                                                         batchedOutType,
+                                                         resultMatrix,
                                                          SmallVector<ReassociationIndices> {
                                                            {0, 1},
                                                            {2}
           });
-    Value result = ONNXTransposeOp::create(rewriter, loc, outType, gemmExpanded, rewriter.getI64ArrayAttr({0, 2, 1}));
+          spatial::SpatYieldOp::create(rewriter, loc, expanded);
+        });
+      batchResults.push_back(batchResultCompute.getResult(0));
+    }
 
+    Value result = concatValues(batchResults, /*axis=*/0, rewriter, loc);
     rewriter.replaceOp(matmulOp, result);
     return success();
   }
@@ -106,7 +247,7 @@ struct MatMulRank3ToGemm : OpRewritePattern<ONNXMatMulOp> {
 } // namespace
 
 void populateMatMulRewritePatterns(RewritePatternSet& patterns, MLIRContext* ctx) {
-  patterns.insert<MatMulRank3ToGemm>(ctx);
+  patterns.insert<MatMulToGemm>(ctx);
 }
 
 } // namespace onnx_mlir

src/PIM/Conversion/ONNXToSpatial/Patterns/Math/ReduceMean.cpp

@@ -5,8 +5,9 @@
 
 #include <algorithm>
 
-#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Common.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Common/Common.hpp"
-#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Patterns.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/ConversionPatterns.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/HostFoldability.hpp"
 #include "src/Accelerators/PIM/Dialect/Spatial/SpatialOps.hpp"
 #include "src/Dialect/ONNX/ONNXOps.hpp"
 
@@ -81,6 +82,24 @@ createAverageCompute(Value input, RankedTensorType resultType, ConversionPattern
   return computeOp.getResult(0);
 }
 
+static Value concatValues(ValueRange inputs, int64_t axis, ConversionPatternRewriter& rewriter, Location loc) {
+  auto firstType = cast<RankedTensorType>(inputs.front().getType());
+  SmallVector<int64_t> outputShape(firstType.getShape().begin(), firstType.getShape().end());
+  int64_t concatDimSize = 0;
+  for (Value input : inputs)
+    concatDimSize += cast<RankedTensorType>(input.getType()).getDimSize(axis);
+  outputShape[axis] = concatDimSize;
+  auto resultType = RankedTensorType::get(outputShape, firstType.getElementType(), firstType.getEncoding());
+
+  if (llvm::all_of(inputs, isHostFoldableValue))
+    return createSpatConcat(rewriter, loc, axis, inputs);
+
+  auto concatCompute = createSpatCompute(rewriter, loc, TypeRange {resultType}, {}, inputs, [&](ValueRange args) {
+    spatial::SpatYieldOp::create(rewriter, loc, createSpatConcat(rewriter, loc, axis, args));
+  });
+  return concatCompute.getResult(0);
+}
+
 static Value buildReduceMeanKeepdims(Value input,
                                      ArrayRef<bool> reducedAxes,
                                      int64_t axis,
@@ -100,8 +119,7 @@ static Value buildReduceMeanKeepdims(Value input,
   for (Value slice : slices)
     reducedSlices.push_back(buildReduceMeanKeepdims(slice, reducedAxes, axis + 1, leafType, rewriter, loc));
 
-  return reducedSlices.size() == 1 ? reducedSlices.front()
+  return concatValues(reducedSlices, axis, rewriter, loc);
-                                   : tensor::ConcatOp::create(rewriter, loc, axis, reducedSlices).getResult();
 }
 
 static Value squeezeReducedAxes(Value keepdimsValue,
@@ -116,9 +134,16 @@ static Value squeezeReducedAxes(Value keepdimsValue,
     return tensor::FromElementsOp::create(rewriter, loc, resultType, ValueRange {element});
   }
 
-  return tensor::CollapseShapeOp::create(
+  auto reassociation = buildCollapseReassociation(reducedAxes);
-           rewriter, loc, resultType, keepdimsValue, buildCollapseReassociation(reducedAxes))
+  if (isHostFoldableValue(keepdimsValue))
-    .getResult();
+    return tensor::CollapseShapeOp::create(rewriter, loc, resultType, keepdimsValue, reassociation).getResult();
+
+  auto squeezeCompute =
+    createSpatCompute<1>(rewriter, loc, TypeRange {resultType}, {}, ValueRange {keepdimsValue}, [&](Value input) {
+      Value collapsed = tensor::CollapseShapeOp::create(rewriter, loc, resultType, input, reassociation);
+      spatial::SpatYieldOp::create(rewriter, loc, collapsed);
+    });
+  return squeezeCompute.getResult(0);
 }
 
 struct ReduceMeanToSpatialCompute : OpConversionPattern<ONNXReduceMeanV13Op> {

src/PIM/Conversion/ONNXToSpatial/Patterns/NN/Pool.cpp

@@ -1,18 +1,20 @@
 #include "mlir/Dialect/Arith/IR/Arith.h"
+#include "mlir/Dialect/SCF/IR/SCF.h"
 #include "mlir/Dialect/Tensor/IR/Tensor.h"
 #include "mlir/IR/BuiltinAttributes.h"
 #include "mlir/IR/BuiltinTypes.h"
 
 #include "llvm/ADT/SmallVector.h"
+#include "llvm/ADT/APFloat.h"
+#include "llvm/ADT/APInt.h"
 
 #include <algorithm>
-#include <cassert>
 #include <optional>
 #include <type_traits>
 
 #include "src/Accelerators/PIM/Common/PimCommon.hpp"
 #include "src/Accelerators/PIM/Compiler/PimCompilerOptions.hpp"
-#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Common.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Common/Common.hpp"
 #include "src/Accelerators/PIM/Dialect/Spatial/SpatialOps.hpp"
 #include "src/Dialect/ONNX/ONNXOps.hpp"
 
@@ -31,13 +33,6 @@ static int64_t getOptionalI64(std::optional<ArrayAttrT> arrayAttr, size_t index,
   return arrayAttr ? getI64(*arrayAttr, index) : defaultValue;
 }
 
-static Value concatAlongAxis(ConversionPatternRewriter& rewriter, Location loc, int64_t axis, ArrayRef<Value> values) {
-  assert(!values.empty() && "Expected at least one value to concatenate.");
-  if (values.size() == 1)
-    return values.front();
-  return tensor::ConcatOp::create(rewriter, loc, axis, values);
-}
-
 static Value materializeContiguousTile(ConversionPatternRewriter& rewriter, Location loc, Value tile) {
   auto tileType = cast<RankedTensorType>(tile.getType());
   Value empty = tensor::EmptyOp::create(rewriter, loc, tileType.getShape(), tileType.getElementType());
@@ -52,27 +47,126 @@ static Value materializeContiguousTile(ConversionPatternRewriter& rewriter, Loca
   return tensor::InsertSliceOp::create(rewriter, loc, tile, empty, offsets, sizes, strides);
 }
 
-template <typename ReduceOp>
+static Value createPoolFillElement(
-static Value reduceWindowValues(ConversionPatternRewriter& rewriter, Location loc, ArrayRef<Value> windowValues) {
+  ConversionPatternRewriter& rewriter, Location loc, Type elementType, bool useMinimumValue) {
-  assert(!windowValues.empty() && "Expected at least one pool window value.");
+  if (!useMinimumValue)
+    return arith::ConstantOp::create(rewriter, loc, elementType, rewriter.getZeroAttr(elementType));
 
-  Value reduced = windowValues.front();
+  if (auto floatType = dyn_cast<FloatType>(elementType)) {
-  for (Value value : windowValues.drop_front())
+    auto minValue = llvm::APFloat::getInf(floatType.getFloatSemantics(), /*Negative=*/true);
-    reduced = ReduceOp::create(rewriter, loc, reduced.getType(), reduced, value);
+    return arith::ConstantOp::create(rewriter, loc, elementType, rewriter.getFloatAttr(floatType, minValue));
-  return reduced;
   }
 
-static Value
+  if (auto integerType = dyn_cast<IntegerType>(elementType)) {
-scaleAverageWindow(ConversionPatternRewriter& rewriter, Location loc, Value reducedWindow, int64_t divisor) {
+    auto minValue = llvm::APInt::getSignedMinValue(integerType.getWidth());
-  assert(divisor > 0 && "AveragePool divisor must be positive.");
+    return arith::ConstantOp::create(rewriter, loc, elementType, rewriter.getIntegerAttr(integerType, minValue));
-  if (divisor == 1)
+  }
-    return reducedWindow;
 
-  auto tileType = cast<RankedTensorType>(reducedWindow.getType());
+  llvm_unreachable("unsupported pool element type");
-  double scale = 1.0 / static_cast<double>(divisor);
+}
-  auto scaleAttr = DenseElementsAttr::get(tileType, rewriter.getFloatAttr(tileType.getElementType(), scale));
+
-  Value scaleTensor = arith::ConstantOp::create(rewriter, loc, tileType, scaleAttr);
+static Value createPoolFillTensor(
-  return spatial::SpatVMulOp::create(rewriter, loc, tileType, reducedWindow, scaleTensor);
+  ConversionPatternRewriter& rewriter, Location loc, RankedTensorType tensorType, bool useMinimumValue) {
+  auto fillElement = createPoolFillElement(rewriter, loc, tensorType.getElementType(), useMinimumValue);
+  return tensor::SplatOp::create(rewriter, loc, tensorType, fillElement);
+}
+
+template <typename PoolOp>
+static Value createPaddedPoolInput(ConversionPatternRewriter& rewriter,
+                                   Location loc,
+                                   PoolOp poolOp,
+                                   Value input,
+                                   RankedTensorType inputType,
+                                   int64_t padTop,
+                                   int64_t padLeft,
+                                   int64_t padBottom,
+                                   int64_t padRight) {
+  if (padTop == 0 && padLeft == 0 && padBottom == 0 && padRight == 0)
+    return input;
+
+  auto paddedType = RankedTensorType::get({inputType.getDimSize(0),
+                                           inputType.getDimSize(1),
+                                           inputType.getDimSize(2) + padTop + padBottom,
+                                           inputType.getDimSize(3) + padLeft + padRight},
+                                          inputType.getElementType(),
+                                          inputType.getEncoding());
+  SmallVector<OpFoldResult> lowPads = {rewriter.getIndexAttr(0),
+                                       rewriter.getIndexAttr(0),
+                                       rewriter.getIndexAttr(padTop),
+                                       rewriter.getIndexAttr(padLeft)};
+  SmallVector<OpFoldResult> highPads = {rewriter.getIndexAttr(0),
+                                        rewriter.getIndexAttr(0),
+                                        rewriter.getIndexAttr(padBottom),
+                                        rewriter.getIndexAttr(padRight)};
+  auto padOp = tensor::PadOp::create(rewriter, loc, paddedType, input, lowPads, highPads);
+  auto* padBlock = new Block();
+  for (int index = 0; index < paddedType.getRank(); ++index)
+    padBlock->addArgument(rewriter.getIndexType(), loc);
+  padOp.getRegion().push_back(padBlock);
+  rewriter.setInsertionPointToStart(padBlock);
+  Value padValue = createPoolFillElement(
+    rewriter, loc, inputType.getElementType(), std::is_same_v<PoolOp, ONNXMaxPoolSingleOutOp>);
+  tensor::YieldOp::create(rewriter, loc, padValue);
+  rewriter.setInsertionPointAfter(padOp);
+  return padOp.getResult();
+}
+
+static FailureOr<Value> createAverageScaleTensor(ConversionPatternRewriter& rewriter,
+                                                 Location loc,
+                                                 Operation* op,
+                                                 RankedTensorType outType,
+                                                 int64_t channels,
+                                                 int64_t inputHeight,
+                                                 int64_t inputWidth,
+                                                 int64_t outputHeight,
+                                                 int64_t outputWidth,
+                                                 int64_t kernelHeight,
+                                                 int64_t kernelWidth,
+                                                 int64_t strideHeight,
+                                                 int64_t strideWidth,
+                                                 int64_t dilationHeight,
+                                                 int64_t dilationWidth,
+                                                 int64_t padTop,
+                                                 int64_t padLeft,
+                                                 bool countIncludePad) {
+  auto elemType = dyn_cast<FloatType>(outType.getElementType());
+  if (!elemType) {
+    op->emitOpError("AveragePool lowering requires a floating-point element type");
+    return failure();
+  }
+
+  auto scaleType = RankedTensorType::get({1, channels, outputHeight, outputWidth}, elemType, outType.getEncoding());
+  SmallVector<Attribute> scaleValues;
+  scaleValues.reserve(static_cast<size_t>(channels * outputHeight * outputWidth));
+  for (int64_t channel = 0; channel < channels; ++channel) {
+    (void) channel;
+    for (int64_t outH = 0; outH < outputHeight; ++outH) {
+      for (int64_t outW = 0; outW < outputWidth; ++outW) {
+        int64_t validCount = 0;
+        for (int64_t kernelH = 0; kernelH < kernelHeight; ++kernelH) {
+          const int64_t inH = outH * strideHeight + kernelH * dilationHeight - padTop;
+          if (inH < 0 || inH >= inputHeight)
+            continue;
+          for (int64_t kernelW = 0; kernelW < kernelWidth; ++kernelW) {
+            const int64_t inW = outW * strideWidth + kernelW * dilationWidth - padLeft;
+            if (inW < 0 || inW >= inputWidth)
+              continue;
+            ++validCount;
+          }
+        }
+
+        const int64_t divisor = countIncludePad ? kernelHeight * kernelWidth : validCount;
+        if (divisor <= 0) {
+          op->emitOpError("AveragePool divisor must be positive");
+          return failure();
+        }
+        scaleValues.push_back(rewriter.getFloatAttr(elemType, 1.0 / static_cast<double>(divisor)));
+      }
+    }
+  }
+
+  auto scaleAttr = DenseElementsAttr::get(scaleType, scaleValues);
+  return arith::ConstantOp::create(rewriter, loc, scaleType, scaleAttr).getResult();
 }
 
 template <typename PoolOp>
@@ -150,49 +244,90 @@ struct PoolToSpatialComputeBase : public OpConversionPattern<PoolOp> {
       }
     }
 
-    (void) padBottom;
-    (void) padRight;
-
     const int64_t xbarSize = static_cast<int64_t>(crossbarSize.getValue());
     const int64_t channelTileCount = (channels + xbarSize - 1) / xbarSize;
+    const int64_t outputPatchCount = batchSize * outputHeight * outputWidth;
+    const bool countIncludePad = [&]() {
+      if constexpr (std::is_same_v<PoolOp, ONNXAveragePoolOp>)
+        return poolOp.getCountIncludePad() == 1;
+      return true;
+    }();
+    Value averageScaleTensor;
+    if constexpr (std::is_same_v<PoolOp, ONNXAveragePoolOp>) {
+      auto maybeAverageScaleTensor = createAverageScaleTensor(rewriter,
+                                                              loc,
+                                                              poolOp,
+                                                              outType,
+                                                              channels,
+                                                              inputHeight,
+                                                              inputWidth,
+                                                              outputHeight,
+                                                              outputWidth,
+                                                              kernelHeight,
+                                                              kernelWidth,
+                                                              strideHeight,
+                                                              strideWidth,
+                                                              dilationHeight,
+                                                              dilationWidth,
+                                                              padTop,
+                                                              padLeft,
+                                                              countIncludePad);
+      if (failed(maybeAverageScaleTensor))
+        return failure();
+      averageScaleTensor = *maybeAverageScaleTensor;
+    }
     constexpr size_t numInputs = 1;
     auto computeOp =
       createSpatCompute<numInputs>(rewriter, loc, outType, {}, ValueRange {x}, [&](Value xArg) -> LogicalResult {
-        SmallVector<Value> batchResults;
+        Value paddedInput = createPaddedPoolInput(rewriter, loc, poolOp, xArg, xType, padTop, padLeft, padBottom, padRight);
-        batchResults.reserve(batchSize);
+        Value pooledOutputInit = tensor::EmptyOp::create(rewriter, loc, outType.getShape(), outType.getElementType());
 
-        for (int64_t batch = 0; batch < batchSize; ++batch) {
+        Value c0 = arith::ConstantIndexOp::create(rewriter, loc, 0);
-          SmallVector<Value> rows;
+        Value c1 = arith::ConstantIndexOp::create(rewriter, loc, 1);
-          rows.reserve(outputHeight);
+        Value cOutputPatchCount = arith::ConstantIndexOp::create(rewriter, loc, outputPatchCount);
+        Value cOutputPixelsPerBatch = arith::ConstantIndexOp::create(rewriter, loc, outputHeight * outputWidth);
+        Value cOutputWidth = arith::ConstantIndexOp::create(rewriter, loc, outputWidth);
+        Value cStrideHeight = arith::ConstantIndexOp::create(rewriter, loc, strideHeight);
+        Value cStrideWidth = arith::ConstantIndexOp::create(rewriter, loc, strideWidth);
 
-          for (int64_t outH = 0; outH < outputHeight; ++outH) {
+        auto outputLoop = scf::ForOp::create(rewriter, loc, c0, cOutputPatchCount, c1, ValueRange {pooledOutputInit});
-            SmallVector<Value> rowPixels;
+        rewriter.setInsertionPointToStart(outputLoop.getBody());
-            rowPixels.reserve(outputWidth);
 
-            for (int64_t outW = 0; outW < outputWidth; ++outW) {
+        Value outputPatchIndex = outputLoop.getInductionVar();
-              SmallVector<Value> outputChannelTiles;
+        Value pooledOutputAcc = outputLoop.getRegionIterArgs().front();
-              outputChannelTiles.reserve(channelTileCount);
 
+        Value batchIndex = arith::DivUIOp::create(rewriter, loc, outputPatchIndex, cOutputPixelsPerBatch);
+        Value batchPatchIndex = arith::RemUIOp::create(rewriter, loc, outputPatchIndex, cOutputPixelsPerBatch);
+        Value outHeightIndex = arith::DivUIOp::create(rewriter, loc, batchPatchIndex, cOutputWidth);
+        Value outWidthIndex = arith::RemUIOp::create(rewriter, loc, batchPatchIndex, cOutputWidth);
+        Value windowBaseH = arith::MulIOp::create(rewriter, loc, outHeightIndex, cStrideHeight);
+        Value windowBaseW = arith::MulIOp::create(rewriter, loc, outWidthIndex, cStrideWidth);
+
+        Value updatedOutput = pooledOutputAcc;
         for (int64_t channelTile = 0; channelTile < channelTileCount; ++channelTile) {
           const int64_t tileChannels = std::min<int64_t>(xbarSize, channels - channelTile * xbarSize);
           auto tileType = RankedTensorType::get({1, tileChannels, 1, 1}, outType.getElementType());
+          Value reducedWindow = createPoolFillTensor(
+            rewriter, loc, tileType, std::is_same_v<PoolOp, ONNXMaxPoolSingleOutOp>);
 
-                SmallVector<Value> windowValues;
-                windowValues.reserve(kernelHeight * kernelWidth);
           for (int64_t kernelH = 0; kernelH < kernelHeight; ++kernelH) {
-                  const int64_t inH = outH * strideHeight + kernelH * dilationHeight - padTop;
+            Value paddedInH = windowBaseH;
-                  if (inH < 0 || inH >= inputHeight)
+            if (kernelH * dilationHeight != 0) {
-                    continue;
+              Value kernelHOffset = arith::ConstantIndexOp::create(rewriter, loc, kernelH * dilationHeight);
+              paddedInH = arith::AddIOp::create(rewriter, loc, paddedInH, kernelHOffset);
+            }
 
             for (int64_t kernelW = 0; kernelW < kernelWidth; ++kernelW) {
-                    const int64_t inW = outW * strideWidth + kernelW * dilationWidth - padLeft;
+              Value paddedInW = windowBaseW;
-                    if (inW < 0 || inW >= inputWidth)
+              if (kernelW * dilationWidth != 0) {
-                      continue;
+                Value kernelWOffset = arith::ConstantIndexOp::create(rewriter, loc, kernelW * dilationWidth);
+                paddedInW = arith::AddIOp::create(rewriter, loc, paddedInW, kernelWOffset);
+              }
 
-                    SmallVector<OpFoldResult> offsets = {rewriter.getIndexAttr(batch),
+              SmallVector<OpFoldResult> offsets = {batchIndex,
                                                    rewriter.getIndexAttr(channelTile * xbarSize),
-                                                         rewriter.getIndexAttr(inH),
+                                                   paddedInH,
-                                                         rewriter.getIndexAttr(inW)};
+                                                   paddedInW};
               SmallVector<OpFoldResult> sizes = {rewriter.getIndexAttr(1),
                                                  rewriter.getIndexAttr(tileChannels),
                                                  rewriter.getIndexAttr(1),
@@ -202,37 +337,51 @@ struct PoolToSpatialComputeBase : public OpConversionPattern<PoolOp> {
                                                    rewriter.getIndexAttr(1),
                                                    rewriter.getIndexAttr(1)};
               Value windowValue =
-                      tensor::ExtractSliceOp::create(rewriter, loc, tileType, xArg, offsets, sizes, strides);
+                tensor::ExtractSliceOp::create(rewriter, loc, tileType, paddedInput, offsets, sizes, strides);
               windowValue = materializeContiguousTile(rewriter, loc, windowValue);
-                    windowValues.push_back(windowValue);
+              reducedWindow = ReduceOp::create(rewriter, loc, tileType, reducedWindow, windowValue);
             }
           }
 
-                if (windowValues.empty())
-                  return rewriter.notifyMatchFailure(poolOp, "pool window resolved to zero valid elements.");
-
-                Value reducedWindow = reduceWindowValues<ReduceOp>(rewriter, loc, windowValues);
           if constexpr (std::is_same_v<PoolOp, ONNXAveragePoolOp>) {
-                  const bool countIncludePad = poolOp.getCountIncludePad() == 1;
+            SmallVector<OpFoldResult> scaleOffsets = {rewriter.getIndexAttr(0),
-                  const int64_t divisor =
+                                                      rewriter.getIndexAttr(channelTile * xbarSize),
-                    countIncludePad ? kernelHeight * kernelWidth : static_cast<int64_t>(windowValues.size());
+                                                      outHeightIndex,
-                  reducedWindow = scaleAverageWindow(rewriter, loc, reducedWindow, divisor);
+                                                      outWidthIndex};
+            SmallVector<OpFoldResult> scaleSizes = {rewriter.getIndexAttr(1),
+                                                    rewriter.getIndexAttr(tileChannels),
+                                                    rewriter.getIndexAttr(1),
+                                                    rewriter.getIndexAttr(1)};
+            SmallVector<OpFoldResult> scaleStrides = {rewriter.getIndexAttr(1),
+                                                      rewriter.getIndexAttr(1),
+                                                      rewriter.getIndexAttr(1),
+                                                      rewriter.getIndexAttr(1)};
+            Value scaleSlice = tensor::ExtractSliceOp::create(
+              rewriter, loc, tileType, averageScaleTensor, scaleOffsets, scaleSizes, scaleStrides);
+            scaleSlice = materializeContiguousTile(rewriter, loc, scaleSlice);
+            reducedWindow = spatial::SpatVMulOp::create(rewriter, loc, tileType, reducedWindow, scaleSlice);
           }
 
-                outputChannelTiles.push_back(reducedWindow);
+          SmallVector<OpFoldResult> outputOffsets = {batchIndex,
+                                                     rewriter.getIndexAttr(channelTile * xbarSize),
+                                                     outHeightIndex,
+                                                     outWidthIndex};
+          SmallVector<OpFoldResult> outputSizes = {rewriter.getIndexAttr(1),
+                                                   rewriter.getIndexAttr(tileChannels),
+                                                   rewriter.getIndexAttr(1),
+                                                   rewriter.getIndexAttr(1)};
+          SmallVector<OpFoldResult> outputStrides = {rewriter.getIndexAttr(1),
+                                                     rewriter.getIndexAttr(1),
+                                                     rewriter.getIndexAttr(1),
+                                                     rewriter.getIndexAttr(1)};
+          updatedOutput = tensor::InsertSliceOp::create(
+            rewriter, loc, reducedWindow, updatedOutput, outputOffsets, outputSizes, outputStrides);
         }
 
-              rowPixels.push_back(concatAlongAxis(rewriter, loc, /*axis=*/1, outputChannelTiles));
+        scf::YieldOp::create(rewriter, loc, updatedOutput);
-            }
 
-            rows.push_back(concatAlongAxis(rewriter, loc, /*axis=*/3, rowPixels));
+        rewriter.setInsertionPointAfter(outputLoop);
-          }
+        spatial::SpatYieldOp::create(rewriter, loc, outputLoop.getResult(0));
-
-          batchResults.push_back(concatAlongAxis(rewriter, loc, /*axis=*/2, rows));
-        }
-
-        Value pooledOutput = concatAlongAxis(rewriter, loc, /*axis=*/0, batchResults);
-        spatial::SpatYieldOp::create(rewriter, loc, pooledOutput);
         return success();
       });
     if (failed(computeOp))

src/PIM/Conversion/ONNXToSpatial/Patterns/NN/Relu.cpp

@@ -1,6 +1,6 @@
 #include "mlir/Transforms/DialectConversion.h"
 
-#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Common.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Common/Common.hpp"
 #include "src/Accelerators/PIM/Dialect/Spatial/SpatialOps.hpp"
 #include "src/Dialect/ONNX/ONNXOps.hpp"
 

src/PIM/Conversion/ONNXToSpatial/Patterns/NN/Sigmoid.cpp

@@ -1,6 +1,6 @@
 #include "mlir/Transforms/DialectConversion.h"
 
-#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Common.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Common/Common.hpp"
 #include "src/Accelerators/PIM/Dialect/Spatial/SpatialOps.hpp"
 #include "src/Dialect/ONNX/ONNXOps.hpp"
 

src/PIM/Conversion/ONNXToSpatial/Patterns/NN/Softmax.cpp

@@ -1,8 +1,9 @@
 #include "mlir/Dialect/Tensor/IR/Tensor.h"
 #include "mlir/Transforms/DialectConversion.h"
 
-#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Common.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Common/Common.hpp"
-#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Patterns.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/ConversionPatterns.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/HostFoldability.hpp"
 #include "src/Accelerators/PIM/Dialect/Spatial/SpatialOps.hpp"
 #include "src/Dialect/ONNX/ONNXOps.hpp"
 
@@ -32,6 +33,24 @@ static Value createSoftmaxCompute(Value input, ConversionPatternRewriter& rewrit
   return computeOp.getResult(0);
 }
 
+static Value concatValues(ValueRange inputs, int64_t axis, ConversionPatternRewriter& rewriter, Location loc) {
+  auto firstType = cast<RankedTensorType>(inputs.front().getType());
+  SmallVector<int64_t> outputShape(firstType.getShape().begin(), firstType.getShape().end());
+  int64_t concatDimSize = 0;
+  for (Value input : inputs)
+    concatDimSize += cast<RankedTensorType>(input.getType()).getDimSize(axis);
+  outputShape[axis] = concatDimSize;
+  auto resultType = RankedTensorType::get(outputShape, firstType.getElementType(), firstType.getEncoding());
+
+  if (llvm::all_of(inputs, isHostFoldableValue))
+    return createSpatConcat(rewriter, loc, axis, inputs);
+
+  auto concatCompute = createSpatCompute(rewriter, loc, TypeRange {resultType}, {}, inputs, [&](ValueRange args) {
+    spatial::SpatYieldOp::create(rewriter, loc, createSpatConcat(rewriter, loc, axis, args));
+  });
+  return concatCompute.getResult(0);
+}
+
 static Value
 buildSoftmax(Value input, int64_t softmaxAxis, int64_t axis, ConversionPatternRewriter& rewriter, Location loc) {
   auto inputType = cast<RankedTensorType>(input.getType());
@@ -47,8 +66,7 @@ buildSoftmax(Value input, int64_t softmaxAxis, int64_t axis, ConversionPatternRe
   for (Value slice : slices)
     rebuiltSlices.push_back(buildSoftmax(slice, softmaxAxis, axis + 1, rewriter, loc));
 
-  return rebuiltSlices.size() == 1 ? rebuiltSlices.front()
+  return concatValues(rebuiltSlices, axis, rewriter, loc);
-                                   : tensor::ConcatOp::create(rewriter, loc, axis, rebuiltSlices).getResult();
 }
 
 struct SoftmaxToSpatialCompute : OpConversionPattern<ONNXSoftmaxOp> {
@@ -93,8 +111,13 @@ struct SoftmaxToSpatialCompute : OpConversionPattern<ONNXSoftmaxOp> {
       Value transposedInput = preTransposeCompute.getResult(0);
       Value transposedResult = buildSoftmax(
         transposedInput, /*softmaxAxis=*/inputType.getRank() - 1, /*axis=*/0, rewriter, softmaxOp.getLoc());
-      result = ONNXTransposeOp::create(
+      auto postTransposeCompute =
-        rewriter, softmaxOp.getLoc(), inputType, transposedResult, rewriter.getI64ArrayAttr(inversePermutation));
+        createSpatCompute<1>(rewriter, softmaxOp.getLoc(), TypeRange {inputType}, {}, transposedResult, [&](Value x) {
+          Value transposed = ONNXTransposeOp::create(
+            rewriter, softmaxOp.getLoc(), inputType, x, rewriter.getI64ArrayAttr(inversePermutation));
+          spatial::SpatYieldOp::create(rewriter, softmaxOp.getLoc(), transposed);
+        });
+      result = postTransposeCompute.getResult(0);
     }
 
     rewriter.replaceOp(softmaxOp, result);

src/PIM/Conversion/ONNXToSpatial/Patterns/Tensor/Concat.cpp

@@ -1,7 +1,10 @@
 #include "mlir/Dialect/Tensor/IR/Tensor.h"
 #include "mlir/IR/PatternMatch.h"
 
-#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Common.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Common/Common.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Common/ComputeRegionBuilder.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/HostFoldability.hpp"
+#include "src/Accelerators/PIM/Dialect/Spatial/SpatialOps.hpp"
 #include "src/Dialect/ONNX/ONNXOps.hpp"
 
 using namespace mlir;
@@ -17,7 +20,17 @@ struct Concat : public OpConversionPattern<ONNXConcatOp> {
     auto inputs = adaptor.getInputs();
     int64_t axis = adaptor.getAxis();
 
-    rewriter.replaceOpWithNewOp<tensor::ConcatOp>(maxpoolOp, axis, inputs);
+    if (llvm::all_of(inputs, isHostFoldableValue)) {
+      rewriter.replaceOp(maxpoolOp, createSpatConcat(rewriter, maxpoolOp.getLoc(), axis, inputs));
+      return success();
+    }
+
+    auto computeOp = createSpatCompute(
+      rewriter, maxpoolOp.getLoc(), TypeRange {maxpoolOp.getResult().getType()}, {}, inputs, [&](ValueRange args) {
+        spatial::SpatYieldOp::create(
+          rewriter, maxpoolOp.getLoc(), createSpatConcat(rewriter, maxpoolOp.getLoc(), axis, args));
+      });
+    rewriter.replaceOp(maxpoolOp, computeOp.getResults());
 
     return success();
   }

src/PIM/Conversion/ONNXToSpatial/Patterns/Tensor/Gather.cpp

@@ -5,8 +5,8 @@
 
 #include "llvm/ADT/SmallVector.h"
 
-#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Common.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Common/Common.hpp"
-#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Patterns.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/ConversionPatterns.hpp"
 #include "src/Accelerators/PIM/Dialect/Spatial/SpatialOps.hpp"
 #include "src/Dialect/ONNX/ONNXOps.hpp"
 
@@ -49,7 +49,7 @@ static Value concatGatherSlices(Value data,
   }
   if (slices.empty())
     return {};
-  return slices.size() == 1 ? slices.front() : tensor::ConcatOp::create(rewriter, loc, axis, slices).getResult();
+  return createSpatConcat(rewriter, loc, axis, slices);
 }
 
 static Value addLeadingGatherDim(Value value, int64_t axis, ConversionPatternRewriter& rewriter, Location loc) {
@@ -130,9 +130,7 @@ struct Gather : OpConversionPattern<ONNXGatherOp> {
                                    return failure();
                                  rows.push_back(addLeadingGatherDim(gatheredRow, axis, rewriter, loc));
                                }
-                               result = rows.size() == 1
+                               result = createSpatConcat(rewriter, loc, /*axis=*/axis, rows);
-                                        ? rows.front()
-                                        : tensor::ConcatOp::create(rewriter, loc, /*axis=*/axis, rows).getResult();
                              }
                              else {
                                return failure();

src/PIM/Conversion/ONNXToSpatial/Patterns/Tensor/Reshape.cpp

@@ -3,7 +3,10 @@
 
 #include "llvm/ADT/SmallVector.h"
 
-#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Patterns.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Common/Common.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/ConversionPatterns.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/HostFoldability.hpp"
+#include "src/Accelerators/PIM/Dialect/Spatial/SpatialOps.hpp"
 #include "src/Dialect/ONNX/ONNXOps.hpp"
 
 using namespace mlir;
@@ -95,18 +98,33 @@ struct Reshape : OpConversionPattern<ONNXReshapeOp> {
       return success();
     }
 
-    SmallVector<ReassociationIndices> reassociation;
+    auto replaceWithReshape = [&](auto buildReshape) -> LogicalResult {
-    if (sourceType.getRank() > resultType.getRank()
+      if (isHostFoldableValue(adaptor.getData())) {
-        && inferCollapseReassociation(sourceType.getShape(), resultType.getShape(), reassociation)) {
+        rewriter.replaceOp(reshapeOp, buildReshape(adaptor.getData()));
-      rewriter.replaceOpWithNewOp<tensor::CollapseShapeOp>(reshapeOp, resultType, adaptor.getData(), reassociation);
         return success();
       }
 
-    if (sourceType.getRank() < resultType.getRank()
+      auto computeOp = createSpatCompute<1>(
-        && inferExpandReassociation(sourceType.getShape(), resultType.getShape(), reassociation)) {
+        rewriter, reshapeOp.getLoc(), TypeRange {resultType}, {}, adaptor.getData(), [&](Value data) {
-      rewriter.replaceOpWithNewOp<tensor::ExpandShapeOp>(reshapeOp, resultType, adaptor.getData(), reassociation);
+          Value reshaped = buildReshape(data);
+          spatial::SpatYieldOp::create(rewriter, reshapeOp.getLoc(), reshaped);
+        });
+      rewriter.replaceOp(reshapeOp, computeOp.getResults());
       return success();
-    }
+    };
+
+    SmallVector<ReassociationIndices> reassociation;
+    if (sourceType.getRank() > resultType.getRank()
+        && inferCollapseReassociation(sourceType.getShape(), resultType.getShape(), reassociation))
+      return replaceWithReshape([&](Value data) {
+        return tensor::CollapseShapeOp::create(rewriter, reshapeOp.getLoc(), resultType, data, reassociation);
+      });
+
+    if (sourceType.getRank() < resultType.getRank()
+        && inferExpandReassociation(sourceType.getShape(), resultType.getShape(), reassociation))
+      return replaceWithReshape([&](Value data) {
+        return tensor::ExpandShapeOp::create(rewriter, reshapeOp.getLoc(), resultType, data, reassociation);
+      });
 
     return failure();
   }

src/PIM/Conversion/ONNXToSpatial/Patterns/Tensor/Resize.cpp

@@ -5,8 +5,8 @@
 
 #include <algorithm>
 
-#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Common.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Common/Common.hpp"
-#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Patterns.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/ConversionPatterns.hpp"
 #include "src/Accelerators/PIM/Dialect/Spatial/SpatialOps.hpp"
 #include "src/Dialect/ONNX/ONNXOps.hpp"
 
@@ -50,7 +50,7 @@ static Value buildNearestResize(Value input,
     slices.push_back(buildNearestResize(slice, inputShape, outputShape, axis + 1, rewriter, loc));
   }
 
-  return slices.size() == 1 ? slices.front() : tensor::ConcatOp::create(rewriter, loc, axis, slices).getResult();
+  return createSpatConcat(rewriter, loc, axis, slices);
 }
 
 struct Resize : OpConversionPattern<ONNXResizeOp> {

src/PIM/Conversion/ONNXToSpatial/Patterns/Tensor/Split.cpp

@@ -1,8 +1,10 @@
 #include "mlir/Dialect/Tensor/IR/Tensor.h"
 #include "mlir/Transforms/DialectConversion.h"
 
-#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Common.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Common/Common.hpp"
-#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Patterns.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/ConversionPatterns.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/HostFoldability.hpp"
+#include "src/Accelerators/PIM/Dialect/Spatial/SpatialOps.hpp"
 #include "src/Dialect/ONNX/ONNXOps.hpp"
 
 using namespace mlir;
@@ -23,7 +25,10 @@ static Value extractSliceAt(
     sizes.push_back(rewriter.getIndexAttr(dim));
   offsets[axis] = rewriter.getIndexAttr(offset);
   sizes[axis] = rewriter.getIndexAttr(size);
-  return tensor::ExtractSliceOp::create(rewriter, loc, input, offsets, sizes, strides);
+  SmallVector<int64_t> resultShape(inputType.getShape());
+  resultShape[axis] = size;
+  auto resultType = RankedTensorType::get(resultShape, inputType.getElementType());
+  return tensor::ExtractSliceOp::create(rewriter, loc, resultType, input, offsets, sizes, strides);
 }
 
 struct Split : OpConversionPattern<ONNXSplitOp> {
@@ -44,23 +49,42 @@ struct Split : OpConversionPattern<ONNXSplitOp> {
     outputs.reserve(splitOp.getNumResults());
 
     int64_t offset = 0;
+    SmallVector<RankedTensorType> resultTypes;
+    resultTypes.reserve(splitOp.getNumResults());
+    SmallVector<int64_t> sliceSizes;
+    sliceSizes.reserve(splitOp.getNumResults());
     for (Value result : splitOp.getResults()) {
       auto resultType = dyn_cast<RankedTensorType>(result.getType());
       if (!resultType || !resultType.hasStaticShape())
         return failure();
-      int64_t sliceSize = resultType.getShape()[axis];
+      resultTypes.push_back(resultType);
-      auto computeOp =
+      sliceSizes.push_back(resultType.getShape()[axis]);
-        createSpatCompute<1>(rewriter, splitOp.getLoc(), TypeRange {resultType}, {}, adaptor.getInput(), [&](Value x) {
-          Value output = extractSliceAt(x, axis, offset, sliceSize, rewriter, splitOp.getLoc());
-          spatial::SpatYieldOp::create(rewriter, splitOp.getLoc(), output);
-        });
-      outputs.push_back(computeOp.getResult(0));
-      offset += sliceSize;
     }
 
+    if (isHostFoldableValue(adaptor.getInput())) {
+      for (int64_t sliceSize : sliceSizes) {
+        outputs.push_back(extractSliceAt(adaptor.getInput(), axis, offset, sliceSize, rewriter, splitOp.getLoc()));
+        offset += sliceSize;
+      }
       rewriter.replaceOp(splitOp, outputs);
       return success();
     }
+
+    auto computeOp = createSpatCompute<1>(
+      rewriter, splitOp.getLoc(), TypeRange(splitOp.getResultTypes()), {}, adaptor.getInput(), [&](Value input) {
+        SmallVector<Value> runtimeOutputs;
+        runtimeOutputs.reserve(resultTypes.size());
+        int64_t runtimeOffset = 0;
+        for (int64_t sliceSize : sliceSizes) {
+          runtimeOutputs.push_back(extractSliceAt(input, axis, runtimeOffset, sliceSize, rewriter, splitOp.getLoc()));
+          runtimeOffset += sliceSize;
+        }
+        spatial::SpatYieldOp::create(rewriter, splitOp.getLoc(), runtimeOutputs);
+      });
+
+    rewriter.replaceOp(splitOp, computeOp.getResults());
+    return success();
+  }
 };
 
 } // namespace

src/PIM/Conversion/ONNXToSpatial/PostPatterns.cpp

@@ -0,0 +1,265 @@
+#include "mlir/Dialect/Arith/IR/Arith.h"
+#include "mlir/Dialect/Func/IR/FuncOps.h"
+#include "mlir/Dialect/Tensor/IR/Tensor.h"
+#include "mlir/IR/IRMapping.h"
+#include "mlir/IR/PatternMatch.h"
+
+#include "llvm/ADT/STLExtras.h"
+#include "llvm/ADT/SmallVector.h"
+
+#include "src/Accelerators/PIM/Common/IR/WeightUtils.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Common/WeightMaterialization.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/PostPatterns.hpp"
+#include "src/Accelerators/PIM/Dialect/Spatial/SpatialOps.hpp"
+#include "src/Dialect/ONNX/ONNXOps.hpp"
+
+using namespace mlir;
+
+namespace onnx_mlir {
+
+namespace {
+
+static bool isWeightMaterializationHelperUser(Operation* op) {
+  return isa<tensor::ExtractSliceOp, tensor::ExpandShapeOp, tensor::CollapseShapeOp, ONNXTransposeOp>(op);
+}
+
+static bool canPromoteInputBlockArgument(BlockArgument arg) {
+  return !arg.use_empty() && llvm::all_of(arg.getUsers(), isWeightMaterializationHelperUser);
+}
+
+static bool isDirectConstantValue(Value value) {
+  return isa_and_nonnull<arith::ConstantOp, ONNXConstantOp>(value.getDefiningOp());
+}
+
+// Collapses one-lane batches so later phases do not carry batch-only structure unnecessarily.
+struct FoldSingleLaneComputeBatchPattern : OpRewritePattern<spatial::SpatComputeBatch> {
+  using OpRewritePattern<spatial::SpatComputeBatch>::OpRewritePattern;
+
+  LogicalResult matchAndRewrite(spatial::SpatComputeBatch batchOp, PatternRewriter& rewriter) const override {
+    if (batchOp.getLaneCount() != 1)
+      return rewriter.notifyMatchFailure(batchOp, "requires a single lane");
+
+    auto loc = batchOp.getLoc();
+    rewriter.setInsertionPoint(batchOp);
+    auto computeOp =
+      spatial::SpatCompute::create(rewriter, loc, batchOp.getResultTypes(), batchOp.getWeights(), batchOp.getInputs());
+    computeOp.getProperties().setOperandSegmentSizes(
+      {static_cast<int>(batchOp.getWeights().size()), static_cast<int>(batchOp.getInputs().size())});
+
+    Block& templateBlock = batchOp.getBody().front();
+    SmallVector<Type> blockArgTypes;
+    SmallVector<Location> blockArgLocs;
+    blockArgTypes.reserve(templateBlock.getNumArguments());
+    blockArgLocs.reserve(templateBlock.getNumArguments());
+    for (BlockArgument arg : templateBlock.getArguments()) {
+      blockArgTypes.push_back(arg.getType());
+      blockArgLocs.push_back(loc);
+    }
+
+    auto* newBlock =
+      rewriter.createBlock(&computeOp.getBody(), computeOp.getBody().end(), TypeRange(blockArgTypes), blockArgLocs);
+    IRMapping mapper;
+    for (auto [oldArg, newArg] : llvm::zip(templateBlock.getArguments(), newBlock->getArguments()))
+      mapper.map(oldArg, newArg);
+
+    rewriter.setInsertionPointToEnd(newBlock);
+    for (Operation& op : templateBlock)
+      rewriter.clone(op, mapper);
+
+    batchOp->replaceAllUsesWith(computeOp->getResults());
+    rewriter.eraseOp(batchOp);
+    return success();
+  }
+};
+
+// Promotes foldable helper chains from runtime inputs to weights to avoid artificial compute inputs.
+struct PromoteWeightLikeComputeInputsPattern : OpRewritePattern<spatial::SpatCompute> {
+  using OpRewritePattern<spatial::SpatCompute>::OpRewritePattern;
+
+  LogicalResult matchAndRewrite(spatial::SpatCompute compute, PatternRewriter& rewriter) const override {
+    SmallVector<bool> promoteInput(compute.getInputs().size(), false);
+    bool needsRewrite = false;
+    Block& oldBlock = compute.getBody().front();
+    for (auto [inputIdx, input] : llvm::enumerate(compute.getInputs())) {
+      if (inputIdx >= oldBlock.getNumArguments())
+        continue;
+      if (!isWeightLikeComputeOperand(input))
+        continue;
+      if (isDirectConstantValue(input) && !canPromoteInputBlockArgument(oldBlock.getArgument(inputIdx)))
+        continue;
+      promoteInput[inputIdx] = true;
+      needsRewrite = true;
+    }
+    if (!needsRewrite)
+      return rewriter.notifyMatchFailure(compute, "no weight-like inputs to promote");
+
+    rewriter.setInsertionPointAfter(compute);
+
+    SmallVector<Value> newWeights(compute.getWeights().begin(), compute.getWeights().end());
+    SmallVector<Value> newInputs;
+    SmallVector<Type> newInputTypes;
+    SmallVector<Location> newInputLocs;
+    newWeights.reserve(compute.getWeights().size() + compute.getInputs().size());
+    newInputs.reserve(compute.getInputs().size());
+    newInputTypes.reserve(compute.getInputs().size());
+    newInputLocs.reserve(compute.getInputs().size());
+
+    for (auto [inputIdx, input] : llvm::enumerate(compute.getInputs())) {
+      if (promoteInput[inputIdx]) {
+        newWeights.push_back(input);
+        continue;
+      }
+      newInputs.push_back(input);
+      newInputTypes.push_back(input.getType());
+      newInputLocs.push_back(input.getLoc());
+    }
+
+    auto newCompute =
+      spatial::SpatCompute::create(rewriter, compute.getLoc(), compute.getResultTypes(), newWeights, newInputs);
+    auto* newBlock =
+      rewriter.createBlock(&newCompute.getBody(), newCompute.getBody().end(), newInputTypes, newInputLocs);
+    newCompute.getProperties().setOperandSegmentSizes(
+      {static_cast<int>(newWeights.size()), static_cast<int>(newInputs.size())});
+    rewriter.setInsertionPointToStart(newBlock);
+
+    IRRewriter bodyRewriter(rewriter.getContext());
+    bodyRewriter.setInsertionPointToStart(newBlock);
+
+    IRMapping mapper;
+    size_t newInputIdx = 0;
+    for (auto [oldInputIdx, oldArg] : llvm::enumerate(oldBlock.getArguments())) {
+      if (!promoteInput[oldInputIdx]) {
+        mapper.map(oldArg, newBlock->getArgument(newInputIdx++));
+        continue;
+      }
+
+      auto clonedValue = materializeWeightLikeValueInBlock(compute.getInputs()[oldInputIdx], bodyRewriter, mapper);
+      if (failed(clonedValue))
+        return rewriter.notifyMatchFailure(compute, "failed to materialize promoted weight-like operand");
+      mapper.map(oldArg, *clonedValue);
+    }
+
+    for (Operation& op : oldBlock.without_terminator())
+      rewriter.clone(op, mapper);
+
+    auto oldYield = cast<spatial::SpatYieldOp>(oldBlock.getTerminator());
+    SmallVector<Value> newYieldOperands;
+    newYieldOperands.reserve(oldYield.getOutputs().size());
+    for (Value operand : oldYield.getOutputs()) {
+      auto mapped = mapper.lookupOrNull(operand);
+      newYieldOperands.push_back(mapped ? cast<Value>(mapped) : operand);
+    }
+    spatial::SpatYieldOp::create(rewriter, oldYield.getLoc(), newYieldOperands);
+
+    rewriter.replaceOp(compute, newCompute.getResults());
+    return success();
+  }
+};
+
+// Promotes foldable batch helper chains to weights while preserving compact compute_batch IR.
+struct PromoteWeightLikeComputeBatchInputsPattern : OpRewritePattern<spatial::SpatComputeBatch> {
+  using OpRewritePattern<spatial::SpatComputeBatch>::OpRewritePattern;
+
+  LogicalResult matchAndRewrite(spatial::SpatComputeBatch compute, PatternRewriter& rewriter) const override {
+    SmallVector<bool> promoteInput(compute.getInputs().size(), false);
+    bool needsRewrite = false;
+    Block& oldBlock = compute.getBody().front();
+    for (auto [inputIdx, input] : llvm::enumerate(compute.getInputs())) {
+      if (inputIdx >= oldBlock.getNumArguments())
+        continue;
+      if (!isWeightLikeComputeOperand(input))
+        continue;
+      if (isDirectConstantValue(input) && !canPromoteInputBlockArgument(oldBlock.getArgument(inputIdx)))
+        continue;
+      promoteInput[inputIdx] = true;
+      needsRewrite = true;
+    }
+    if (!needsRewrite)
+      return rewriter.notifyMatchFailure(compute, "no weight-like batch inputs to promote");
+
+    rewriter.setInsertionPointAfter(compute);
+
+    SmallVector<Value> newWeights(compute.getWeights().begin(), compute.getWeights().end());
+    SmallVector<Value> newInputs;
+    SmallVector<Type> newInputTypes;
+    SmallVector<Location> newInputLocs;
+    newWeights.reserve(compute.getWeights().size() + compute.getInputs().size());
+    newInputs.reserve(compute.getInputs().size());
+    newInputTypes.reserve(compute.getInputs().size());
+    newInputLocs.reserve(compute.getInputs().size());
+
+    for (auto [inputIdx, input] : llvm::enumerate(compute.getInputs())) {
+      if (promoteInput[inputIdx]) {
+        newWeights.push_back(input);
+        continue;
+      }
+      newInputs.push_back(input);
+      newInputTypes.push_back(input.getType());
+      newInputLocs.push_back(input.getLoc());
+    }
+
+    auto newCompute =
+      spatial::SpatComputeBatch::create(rewriter,
+                                        compute.getLoc(),
+                                        compute.getResultTypes(),
+                                        rewriter.getI32IntegerAttr(static_cast<int32_t>(compute.getLaneCount())),
+                                        newWeights,
+                                        newInputs);
+    auto* newBlock =
+      rewriter.createBlock(&newCompute.getBody(), newCompute.getBody().end(), newInputTypes, newInputLocs);
+    newCompute.getProperties().setOperandSegmentSizes(
+      {static_cast<int>(newWeights.size()), static_cast<int>(newInputs.size())});
+    rewriter.setInsertionPointToStart(newBlock);
+
+    IRRewriter bodyRewriter(rewriter.getContext());
+    bodyRewriter.setInsertionPointToStart(newBlock);
+
+    IRMapping mapper;
+    size_t newInputIdx = 0;
+    for (auto [oldInputIdx, oldArg] : llvm::enumerate(oldBlock.getArguments())) {
+      if (!promoteInput[oldInputIdx]) {
+        mapper.map(oldArg, newBlock->getArgument(newInputIdx++));
+        continue;
+      }
+
+      auto clonedValue = materializeWeightLikeValueInBlock(compute.getInputs()[oldInputIdx], bodyRewriter, mapper);
+      if (failed(clonedValue))
+        return rewriter.notifyMatchFailure(compute, "failed to materialize promoted batch weight-like operand");
+      mapper.map(oldArg, *clonedValue);
+    }
+
+    for (Operation& op : oldBlock.without_terminator())
+      rewriter.clone(op, mapper);
+
+    auto oldYield = cast<spatial::SpatYieldOp>(oldBlock.getTerminator());
+    SmallVector<Value> newYieldOperands;
+    newYieldOperands.reserve(oldYield.getOutputs().size());
+    for (Value operand : oldYield.getOutputs()) {
+      auto mapped = mapper.lookupOrNull(operand);
+      newYieldOperands.push_back(mapped ? cast<Value>(mapped) : operand);
+    }
+    spatial::SpatYieldOp::create(rewriter, oldYield.getLoc(), newYieldOperands);
+
+    rewriter.replaceOp(compute, newCompute.getResults());
+    return success();
+  }
+};
+
+} // namespace
+
+void populateEarlyPostPatterns(RewritePatternSet& patterns, MLIRContext* ctx) {
+  patterns.add<FoldSingleLaneComputeBatchPattern>(ctx);
+}
+
+void populatePostPatterns(RewritePatternSet& patterns, MLIRContext* ctx) {
+  patterns.add<PromoteWeightLikeComputeInputsPattern, PromoteWeightLikeComputeBatchInputsPattern>(ctx);
+}
+
+void annotateWeightsConstants(func::FuncOp funcOp) {
+  funcOp.walk([&](arith::ConstantOp constantOp) {
+    if (hasOnlySpatialMvmVmmWeightUses(constantOp.getResult()))
+      markWeightAlways(constantOp);
+  });
+}
+
+} // namespace onnx_mlir

src/PIM/Conversion/ONNXToSpatial/PostPatterns.hpp

@@ -0,0 +1,14 @@
+#pragma once
+
+#include "mlir/Dialect/Func/IR/FuncOps.h"
+#include "mlir/IR/MLIRContext.h"
+
+namespace onnx_mlir {
+
+void populateEarlyPostPatterns(mlir::RewritePatternSet& patterns, mlir::MLIRContext* ctx);
+
+void populatePostPatterns(mlir::RewritePatternSet& patterns, mlir::MLIRContext* ctx);
+
+void annotateWeightsConstants(mlir::func::FuncOp funcOp);
+
+} // namespace onnx_mlir

src/PIM/Conversion/ONNXToSpatial/PrePatterns.cpp

@@ -0,0 +1,25 @@
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/Common/Common.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/ConversionPatterns.hpp"
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/PrePatterns.hpp"
+
+using namespace mlir;
+
+namespace onnx_mlir {
+
+namespace {
+
+#include "src/Accelerators/PIM/Conversion/ONNXToSpatial/ONNXToSpatial.hpp.inc"
+
+} // namespace
+
+void populatePrePatterns(mlir::RewritePatternSet& patterns, mlir::MLIRContext* ctx) {
+  patterns.add<onnxToArithConstant>(ctx);
+  patterns.add<convAddToConvWithBiasLeft>(ctx);
+  patterns.add<convAddToConvWithBiasRight>(ctx);
+  patterns.add<matMulAddToGemm>(ctx);
+  patterns.add<matMulToGemm>(ctx);
+  patterns.add<removeFlattenSameShape>(ctx);
+  populateMatMulRewritePatterns(patterns, ctx);
+}
+
+} // namespace onnx_mlir

src/PIM/Conversion/ONNXToSpatial/PrePatterns.hpp

@@ -0,0 +1,10 @@
+#pragma once
+
+#include "mlir/IR/MLIRContext.h"
+#include "mlir/Transforms/DialectConversion.h"
+
+namespace onnx_mlir {
+
+void populatePrePatterns(mlir::RewritePatternSet& patterns, mlir::MLIRContext* ctx);
+
+} // namespace onnx_mlir

src/PIM/Conversion/SpatialToPim/BatchCoreLoweringPatterns.cpp

@@ -0,0 +1,218 @@
+#include "mlir/Dialect/Bufferization/IR/Bufferization.h"
+#include "mlir/Dialect/MemRef/IR/MemRef.h"
+#include "mlir/Dialect/Tensor/IR/Tensor.h"
+#include "mlir/IR/IRMapping.h"
+
+#include "Conversion/ONNXToSpatial/Common/Common.hpp"
+#include "src/Accelerators/PIM/Common/PimCommon.hpp"
+#include "src/Accelerators/PIM/Conversion/SpatialToPim/BatchCoreLoweringPatterns.hpp"
+#include "src/Accelerators/PIM/Conversion/SpatialToPim/Common.hpp"
+#include "src/Accelerators/PIM/Dialect/Pim/PimOps.hpp"
+
+using namespace mlir;
+using namespace onnx_mlir::pim;
+
+namespace onnx_mlir {
+namespace {
+
+static int32_t translateSpatialCoreIdToPimCoreId(size_t spatialCoreId) { return static_cast<int32_t>(spatialCoreId); }
+
+static SmallVector<int32_t> getPimCoreIdsForBatchOp(spatial::SpatComputeBatch computeBatchOp, size_t& fallbackCoreId) {
+  if (auto coreIdsAttr = computeBatchOp->getAttrOfType<DenseI32ArrayAttr>(onnx_mlir::kCoreIdsAttrName))
+    return SmallVector<int32_t>(coreIdsAttr.asArrayRef().begin(), coreIdsAttr.asArrayRef().end());
+
+  SmallVector<int32_t> coreIds;
+  coreIds.reserve(static_cast<size_t>(computeBatchOp.getLaneCount()));
+  for (uint32_t lane = 0; lane < computeBatchOp.getLaneCount(); ++lane)
+    coreIds.push_back(static_cast<int32_t>(fallbackCoreId++));
+  return coreIds;
+}
+
+static void lowerChannelSendTensorBatch(spatial::SpatChannelSendTensorBatchOp sendTensorBatchOp,
+                                        IRMapping& mapper,
+                                        IRRewriter& rewriter) {
+  SmallVector<int32_t> targetCoreIds;
+  targetCoreIds.reserve(sendTensorBatchOp.getTargetCoreIds().size());
+  for (int32_t targetCoreId : sendTensorBatchOp.getTargetCoreIds())
+    targetCoreIds.push_back(translateSpatialCoreIdToPimCoreId(targetCoreId));
+
+  pim::PimSendTensorBatchOp::create(rewriter,
+                                    sendTensorBatchOp.getLoc(),
+                                    mapper.lookup(sendTensorBatchOp.getInput()),
+                                    rewriter.getDenseI32ArrayAttr(targetCoreIds));
+}
+
+static void lowerChannelReceiveTensorBatch(spatial::SpatChannelReceiveTensorBatchOp receiveTensorBatchOp,
+                                           IRMapping& mapper,
+                                           IRRewriter& rewriter) {
+  SmallVector<int32_t> sourceCoreIds;
+  sourceCoreIds.reserve(receiveTensorBatchOp.getSourceCoreIds().size());
+  for (int32_t sourceCoreId : receiveTensorBatchOp.getSourceCoreIds())
+    sourceCoreIds.push_back(translateSpatialCoreIdToPimCoreId(sourceCoreId));
+
+  auto outputType = cast<ShapedType>(receiveTensorBatchOp.getOutput().getType());
+  auto outputBuffer = createEmptyTensorFromShaped(rewriter, receiveTensorBatchOp.getLoc(), outputType);
+  Value received = pim::PimReceiveTensorBatchOp::create(rewriter,
+                                                        receiveTensorBatchOp.getLoc(),
+                                                        outputBuffer.getType(),
+                                                        outputBuffer,
+                                                        rewriter.getDenseI32ArrayAttr(sourceCoreIds))
+                     .getOutput();
+  mapper.map(receiveTensorBatchOp.getOutput(), received);
+}
+
+} // namespace
+
+LogicalResult
+lowerComputeBatchOp(spatial::SpatComputeBatch computeBatchOp, CoreLoweringState& state, IRRewriter& rewriter) {
+  if (computeBatchOp.getNumResults() != 0)
+    return computeBatchOp.emitOpError(
+      "batched Spatial-to-PIM lowering currently requires channelized compute_batch with no results");
+
+  Location loc = computeBatchOp.getLoc();
+  Block& oldBlock = computeBatchOp.getBody().front();
+  auto oldYield = cast<spatial::SpatYieldOp>(oldBlock.getTerminator());
+  if (oldYield.getNumOperands() != 0)
+    return computeBatchOp.emitOpError("batched Spatial-to-PIM lowering currently requires empty spat.yield");
+
+  SmallVector<int32_t> coreIds = getPimCoreIdsForBatchOp(computeBatchOp, state.nextCoreId);
+  SmallVector<Value> batchWeights(computeBatchOp.getWeights().begin(), computeBatchOp.getWeights().end());
+  SmallVector<Value> batchInputs;
+  if (!computeBatchOp.getInputs().empty())
+    batchInputs.append(computeBatchOp.getInputs().begin(), computeBatchOp.getInputs().end());
+
+  rewriter.setInsertionPointAfter(computeBatchOp);
+  auto coreBatchOp = pim::PimCoreBatchOp::create(rewriter,
+                                                 loc,
+                                                 rewriter.getI32IntegerAttr(computeBatchOp.getLaneCount()),
+                                                 ValueRange(batchWeights),
+                                                 ValueRange(batchInputs));
+  coreBatchOp.getProperties().setOperandSegmentSizes(
+    {static_cast<int>(batchWeights.size()), static_cast<int>(batchInputs.size())});
+  coreBatchOp->setAttr(onnx_mlir::kCoreIdsAttrName, rewriter.getDenseI32ArrayAttr(coreIds));
+
+  SmallVector<Type> blockArgTypes;
+  SmallVector<Location> blockArgLocs;
+  for (BlockArgument arg : oldBlock.getArguments()) {
+    blockArgTypes.push_back(arg.getType());
+    blockArgLocs.push_back(arg.getLoc());
+  }
+  Block* newBlock =
+    rewriter.createBlock(&coreBatchOp.getBody(), coreBatchOp.getBody().end(), TypeRange(blockArgTypes), blockArgLocs);
+
+  IRMapping mapper;
+  rewriter.setInsertionPointToStart(newBlock);
+  for (auto [oldArg, newArg] : llvm::zip(oldBlock.getArguments(), newBlock->getArguments())) {
+    auto newArgType = cast<ShapedType>(newArg.getType());
+    auto outputBuffer = createEmptyTensorFromShaped(rewriter, loc, newArgType);
+    auto copied = pim::PimMemCopyHostToDevBatchOp::create(rewriter,
+                                                          loc,
+                                                          outputBuffer.getType(),
+                                                          outputBuffer,
+                                                          newArg,
+                                                          rewriter.getI32IntegerAttr(0),
+                                                          rewriter.getI32IntegerAttr(0),
+                                                          getTensorSizeInBytesAttr(rewriter, newArg))
+                    .getOutput();
+    mapper.map(oldArg, copied);
+  }
+
+  auto materializeCapturedTensor = [&](Value capturedTensor) -> Value {
+    if (auto mapped = mapper.lookupOrNull(capturedTensor))
+      return mapped;
+
+    auto capturedType = cast<ShapedType>(capturedTensor.getType());
+    auto outputBuffer = createEmptyTensorFromShaped(rewriter, loc, capturedType);
+    auto copied = pim::PimMemCopyHostToDevBatchOp::create(rewriter,
+                                                          loc,
+                                                          outputBuffer.getType(),
+                                                          outputBuffer,
+                                                          capturedTensor,
+                                                          rewriter.getI32IntegerAttr(0),
+                                                          rewriter.getI32IntegerAttr(0),
+                                                          getTensorSizeInBytesAttr(rewriter, capturedTensor))
+                    .getOutput();
+    mapper.map(capturedTensor, copied);
+    return copied;
+  };
+
+  rewriter.setInsertionPointToEnd(newBlock);
+  for (Operation& op : oldBlock) {
+    if (isa<spatial::SpatYieldOp>(op))
+      continue;
+
+    if (auto sendBatchOp = dyn_cast<spatial::SpatChannelSendBatchOp>(op)) {
+      pim::PimSendBatchOp::create(rewriter,
+                                  loc,
+                                  mapper.lookup(sendBatchOp.getInput()),
+                                  getTensorSizeInBytesAttr(rewriter, mapper.lookup(sendBatchOp.getInput())),
+                                  sendBatchOp.getTargetCoreIdsAttr());
+      continue;
+    }
+
+    if (auto sendTensorBatchOp = dyn_cast<spatial::SpatChannelSendTensorBatchOp>(op)) {
+      lowerChannelSendTensorBatch(sendTensorBatchOp, mapper, rewriter);
+      continue;
+    }
+
+    if (auto receiveBatchOp = dyn_cast<spatial::SpatChannelReceiveBatchOp>(op)) {
+      auto outputType = cast<ShapedType>(receiveBatchOp.getOutput().getType());
+      auto outputBuffer = createEmptyTensorFromShaped(rewriter, loc, outputType);
+      auto received = pim::PimReceiveBatchOp::create(rewriter,
+                                                     loc,
+                                                     outputBuffer.getType(),
+                                                     outputBuffer,
+                                                     getTensorSizeInBytesAttr(rewriter, receiveBatchOp.getOutput()),
+                                                     receiveBatchOp.getSourceCoreIdsAttr())
+                        .getOutput();
+      mapper.map(receiveBatchOp.getOutput(), received);
+      continue;
+    }
+
+    if (auto receiveTensorBatchOp = dyn_cast<spatial::SpatChannelReceiveTensorBatchOp>(op)) {
+      lowerChannelReceiveTensorBatch(receiveTensorBatchOp, mapper, rewriter);
+      continue;
+    }
+
+    if (auto toTensorOp = dyn_cast<bufferization::ToTensorOp>(op)) {
+      if (isa_and_present<memref::GetGlobalOp>(toTensorOp.getBuffer().getDefiningOp())) {
+        Operation* cloned = rewriter.clone(op, mapper);
+        auto clonedTensor = cloned->getResult(0);
+        auto clonedType = cast<ShapedType>(clonedTensor.getType());
+        auto outputBuffer = createEmptyTensorFromShaped(rewriter, loc, clonedType);
+        auto copied = pim::PimMemCopyHostToDevBatchOp::create(rewriter,
+                                                              loc,
+                                                              outputBuffer.getType(),
+                                                              outputBuffer,
+                                                              clonedTensor,
+                                                              rewriter.getI32IntegerAttr(0),
+                                                              rewriter.getI32IntegerAttr(0),
+                                                              getTensorSizeInBytesAttr(rewriter, clonedTensor))
+                        .getOutput();
+        mapper.map(toTensorOp.getResult(), copied);
+        continue;
+      }
+    }
+
+    for (Value operand : op.getOperands()) {
+      if (!isa<TensorType>(operand.getType()) || mapper.contains(operand))
+        continue;
+
+      Operation* definingOp = operand.getDefiningOp();
+      if (definingOp && definingOp->getBlock() == &oldBlock)
+        continue;
+
+      materializeCapturedTensor(operand);
+    }
+
+    Operation* cloned = rewriter.clone(op, mapper);
+    for (auto [originalResult, clonedResult] : llvm::zip(op.getResults(), cloned->getResults()))
+      mapper.map(originalResult, clonedResult);
+  }
+
+  rewriter.setInsertionPointToEnd(newBlock);
+  PimHaltOp::create(rewriter, loc);
+  return success();
+}
+
+} // namespace onnx_mlir

src/PIM/Conversion/SpatialToPim/BatchCoreLoweringPatterns.hpp

@@ -0,0 +1,10 @@
+#pragma once
+
+#include "src/Accelerators/PIM/Conversion/SpatialToPim/CoreLoweringPatterns.hpp"
+
+namespace onnx_mlir {
+
+mlir::LogicalResult
+lowerComputeBatchOp(spatial::SpatComputeBatch computeBatchOp, CoreLoweringState& state, mlir::IRRewriter& rewriter);
+
+} // namespace onnx_mlir

src/PIM/Conversion/SpatialToPim/CMakeLists.txt

@@ -4,7 +4,16 @@ add_public_tablegen_target(SpatialToPimIncGen)
 
 add_pim_library(OMSpatialToPim
   SpatialToPimPass.cpp
+  BatchCoreLoweringPatterns.cpp
+  ChannelLoweringPatterns.cpp
+  Cleanup.cpp
   Common.cpp
+  ComputeLikeRegionUtils.cpp
+  CoreLoweringPatterns.cpp
+  GlobalTensorMaterialization.cpp
+  PhaseVerification.cpp
+  ReturnPathNormalization.cpp
+  TensorPackingPatterns.cpp
 
   EXCLUDE_FROM_OM_LIBS
 

src/PIM/Conversion/SpatialToPim/ChannelLoweringPatterns.cpp

@@ -0,0 +1,136 @@
+#include "mlir/Dialect/Tensor/IR/Tensor.h"
+
+#include "src/Accelerators/PIM/Conversion/SpatialToPim/ChannelLoweringPatterns.hpp"
+#include "src/Accelerators/PIM/Conversion/SpatialToPim/Common.hpp"
+#include "src/Accelerators/PIM/Dialect/Pim/PimOps.hpp"
+#include "src/Accelerators/PIM/Dialect/Spatial/SpatialOps.hpp"
+
+using namespace mlir;
+
+namespace onnx_mlir {
+namespace {
+
+static int32_t toPimCoreId(int32_t spatialCoreId) { return spatialCoreId; }
+
+struct ChannelSendLowering : OpRewritePattern<spatial::SpatChannelSendOp> {
+  using OpRewritePattern::OpRewritePattern;
+
+  LogicalResult matchAndRewrite(spatial::SpatChannelSendOp op, PatternRewriter& rewriter) const override {
+    pim::PimSendOp::create(rewriter,
+                           op.getLoc(),
+                           op.getInput(),
+                           getTensorSizeInBytesAttr(rewriter, op.getInput()),
+                           rewriter.getI32IntegerAttr(toPimCoreId(op.getTargetCoreId())));
+    rewriter.eraseOp(op);
+    return success();
+  }
+};
+
+struct ChannelReceiveLowering : OpRewritePattern<spatial::SpatChannelReceiveOp> {
+  using OpRewritePattern::OpRewritePattern;
+
+  LogicalResult matchAndRewrite(spatial::SpatChannelReceiveOp op, PatternRewriter& rewriter) const override {
+    if (op->use_empty()) {
+      rewriter.eraseOp(op);
+      return success();
+    }
+    auto outputType = cast<ShapedType>(op.getResult().getType());
+    Value outputBuffer =
+      tensor::EmptyOp::create(rewriter, op.getLoc(), outputType.getShape(), outputType.getElementType()).getResult();
+    Value received = pim::PimReceiveOp::create(rewriter,
+                                               op.getLoc(),
+                                               op.getResult().getType(),
+                                               outputBuffer,
+                                               getTensorSizeInBytesAttr(rewriter, op.getResult()),
+                                               rewriter.getI32IntegerAttr(toPimCoreId(op.getSourceCoreId())))
+                       .getOutput();
+    rewriter.replaceOp(op, received);
+    return success();
+  }
+};
+
+struct ChannelSendTensorLowering : OpRewritePattern<spatial::SpatChannelSendTensorOp> {
+  using OpRewritePattern::OpRewritePattern;
+
+  LogicalResult matchAndRewrite(spatial::SpatChannelSendTensorOp op, PatternRewriter& rewriter) const override {
+    SmallVector<int32_t> targetCoreIds;
+    targetCoreIds.reserve(op.getTargetCoreIds().size());
+    for (int32_t targetCoreId : op.getTargetCoreIds())
+      targetCoreIds.push_back(toPimCoreId(targetCoreId));
+    pim::PimSendTensorOp::create(rewriter, op.getLoc(), op.getInput(), rewriter.getDenseI32ArrayAttr(targetCoreIds));
+    rewriter.eraseOp(op);
+    return success();
+  }
+};
+
+struct ChannelReceiveTensorLowering : OpRewritePattern<spatial::SpatChannelReceiveTensorOp> {
+  using OpRewritePattern::OpRewritePattern;
+
+  LogicalResult matchAndRewrite(spatial::SpatChannelReceiveTensorOp op, PatternRewriter& rewriter) const override {
+    SmallVector<int32_t> sourceCoreIds;
+    sourceCoreIds.reserve(op.getSourceCoreIds().size());
+    for (int32_t sourceCoreId : op.getSourceCoreIds())
+      sourceCoreIds.push_back(toPimCoreId(sourceCoreId));
+    auto outputType = cast<ShapedType>(op.getOutput().getType());
+    Value outputBuffer =
+      tensor::EmptyOp::create(rewriter, op.getLoc(), outputType.getShape(), outputType.getElementType()).getResult();
+    Value received =
+      pim::PimReceiveTensorOp::create(
+        rewriter, op.getLoc(), op.getOutput().getType(), outputBuffer, rewriter.getDenseI32ArrayAttr(sourceCoreIds))
+        .getOutput();
+    rewriter.replaceOp(op, received);
+    return success();
+  }
+};
+
+struct ExtractRowsLowering : OpRewritePattern<spatial::SpatExtractRowsOp> {
+  using OpRewritePattern::OpRewritePattern;
+
+  LogicalResult matchAndRewrite(spatial::SpatExtractRowsOp op, PatternRewriter& rewriter) const override {
+    auto inputType = cast<RankedTensorType>(op.getInput().getType());
+    SmallVector<Value> replacements;
+    replacements.reserve(op.getNumResults());
+    for (auto [rowIndex, output] : llvm::enumerate(op.getOutputs())) {
+      auto outputType = cast<RankedTensorType>(output.getType());
+      SmallVector<OpFoldResult> offsets = {
+        rewriter.getIndexAttr(static_cast<int64_t>(rowIndex) * outputType.getDimSize(0)), rewriter.getIndexAttr(0)};
+      SmallVector<OpFoldResult> sizes = {rewriter.getIndexAttr(outputType.getDimSize(0)),
+                                         rewriter.getIndexAttr(inputType.getDimSize(1))};
+      SmallVector<OpFoldResult> strides = {rewriter.getIndexAttr(1), rewriter.getIndexAttr(1)};
+      replacements.push_back(
+        tensor::ExtractSliceOp::create(rewriter, op.getLoc(), outputType, op.getInput(), offsets, sizes, strides)
+          .getResult());
+    }
+    rewriter.replaceOp(op, replacements);
+    return success();
+  }
+};
+
+struct ConcatLowering : OpRewritePattern<spatial::SpatConcatOp> {
+  using OpRewritePattern::OpRewritePattern;
+
+  LogicalResult matchAndRewrite(spatial::SpatConcatOp op, PatternRewriter& rewriter) const override {
+    auto outputType = cast<ShapedType>(op.getOutput().getType());
+    Value outputBuffer =
+      tensor::EmptyOp::create(rewriter, op.getLoc(), outputType.getShape(), outputType.getElementType()).getResult();
+    Value concatenated =
+      pim::PimConcatOp::create(
+        rewriter, op.getLoc(), op.getOutput().getType(), op.getAxisAttr(), op.getInputs(), outputBuffer)
+        .getOutput();
+    rewriter.replaceOp(op, concatenated);
+    return success();
+  }
+};
+
+} // namespace
+
+void populateChannelLoweringPatterns(RewritePatternSet& patterns) {
+  patterns.add<ChannelSendLowering,
+               ChannelReceiveLowering,
+               ChannelSendTensorLowering,
+               ChannelReceiveTensorLowering,
+               ExtractRowsLowering,
+               ConcatLowering>(patterns.getContext());
+}
+
+} // namespace onnx_mlir

src/PIM/Conversion/SpatialToPim/ChannelLoweringPatterns.hpp

@@ -0,0 +1,9 @@
+#pragma once
+
+#include "mlir/IR/PatternMatch.h"
+
+namespace onnx_mlir {
+
+void populateChannelLoweringPatterns(mlir::RewritePatternSet& patterns);
+
+} // namespace onnx_mlir

src/PIM/Conversion/SpatialToPim/Cleanup.cpp

@@ -0,0 +1,42 @@
+#include "llvm/ADT/STLExtras.h"
+
+#include "src/Accelerators/PIM/Conversion/SpatialToPim/Cleanup.hpp"
+
+using namespace mlir;
+
+namespace onnx_mlir {
+
+LogicalResult erasePendingOps(SmallVectorImpl<Operation*>& pendingOps, IRRewriter& rewriter) {
+  while (!pendingOps.empty()) {
+    bool erasedAnyOp = false;
+    for (auto it = pendingOps.begin(); it != pendingOps.end();) {
+      Operation* opToRemove = *it;
+      if (!opToRemove->use_empty()) {
+        ++it;
+        continue;
+      }
+
+      rewriter.eraseOp(opToRemove);
+      it = pendingOps.erase(it);
+      erasedAnyOp = true;
+    }
+
+    if (erasedAnyOp)
+      continue;
+
+    for (Operation* opToRemove : pendingOps) {
+      InFlightDiagnostic diag = opToRemove->emitError("pending Spatial-to-PIM cleanup could not erase operation");
+      diag << "; op has " << llvm::range_size(opToRemove->getUsers()) << " remaining user(s)";
+      for (Operation* user : opToRemove->getUsers()) {
+        bool userPendingRemoval = llvm::is_contained(pendingOps, user);
+        opToRemove->emitRemark() << "remaining user `" << user->getName() << "`"
+                                 << (userPendingRemoval ? " is also pending removal" : " is not pending removal");
+      }
+    }
+    return failure();
+  }
+
+  return success();
+}
+
+} // namespace onnx_mlir

src/PIM/Conversion/SpatialToPim/Cleanup.hpp

@@ -0,0 +1,11 @@
+#pragma once
+
+#include "mlir/IR/Operation.h"
+#include "mlir/IR/PatternMatch.h"
+#include "mlir/Support/LLVM.h"
+
+namespace onnx_mlir {
+
+mlir::LogicalResult erasePendingOps(llvm::SmallVectorImpl<mlir::Operation*>& pendingOps, mlir::IRRewriter& rewriter);
+
+} // namespace onnx_mlir

src/PIM/Conversion/SpatialToPim/Common.cpp

@@ -7,23 +7,12 @@
 #include <cstddef>
 
 #include "Common.hpp"
-#include "src/Accelerators/PIM/Dialect/Pim/PimOps.hpp"
 
 using namespace llvm;
 using namespace mlir;
 
 namespace onnx_mlir {
 
-namespace {
-
-IntegerAttr getRequiredI32Attr(Builder& builder, Operation* op, llvm::StringRef attrName) {
-  auto attr = op->getAttrOfType<IntegerAttr>(attrName);
-  assert(attr && "required precomputed channel attr is missing");
-  return IntegerAttr::get(builder.getI32Type(), attr.getInt());
-}
-
-} // namespace
-
 size_t getSliceActualOffset(tensor::ExtractSliceOp& sliceOp, ShapedType& inputShape) {
   /*
     EXAMPLE RUN:
@@ -74,37 +63,6 @@ IntegerAttr getTensorSizeInBytesAttr(Builder& builder, mlir::Value value) {
   return builder.getI32IntegerAttr(static_cast<int32_t>(getShapedTypeSizeInBytes(cast<ShapedType>(value.getType()))));
 }
 
-IntegerAttr getSpatialChannelSourceCoreIdAttr(Builder& builder, mlir::Value channel) {
-  auto channelNewOp = channel.getDefiningOp<spatial::SpatChannelNewOp>();
-  assert(channelNewOp && "spatial channel value must come from spat.channel_new");
-  return getRequiredI32Attr(builder, channelNewOp, kChannelSourceCoreIdAttrName);
-}
-
-IntegerAttr getSpatialChannelTargetCoreIdAttr(Builder& builder, mlir::Value channel) {
-  auto channelNewOp = channel.getDefiningOp<spatial::SpatChannelNewOp>();
-  assert(channelNewOp && "spatial channel value must come from spat.channel_new");
-  return getRequiredI32Attr(builder, channelNewOp, kChannelTargetCoreIdAttrName);
-}
-
-bool hasSpatialChannelSourceCoreIdAttr(mlir::Value channel) {
-  auto channelNewOp = channel.getDefiningOp<spatial::SpatChannelNewOp>();
-  return channelNewOp && channelNewOp->hasAttr(kChannelSourceCoreIdAttrName);
-}
-
-bool hasSpatialChannelTargetCoreIdAttr(mlir::Value channel) {
-  auto channelNewOp = channel.getDefiningOp<spatial::SpatChannelNewOp>();
-  return channelNewOp && channelNewOp->hasAttr(kChannelTargetCoreIdAttrName);
-}
-
-mlir::Value createPimReceiveFromSpatialChannel(
-  PatternRewriter& rewriter, Location loc, mlir::Value output, mlir::Value channel) {
-  mlir::Value outputBuffer = getBestOutputTensorFromOperandsOrAllocate(rewriter, output.getDefiningOp());
-  auto sizeAttr = getTensorSizeInBytesAttr(rewriter, output);
-  auto sourceCoreIdAttr = getSpatialChannelSourceCoreIdAttr(rewriter, channel);
-  return pim::PimReceiveOp::create(rewriter, loc, outputBuffer.getType(), outputBuffer, sizeAttr, sourceCoreIdAttr)
-    .getOutput();
-}
-
 Operation* getEarliestUserWithinBlock(mlir::Value value) {
   auto users = value.getUsers();
 
@@ -127,15 +85,26 @@ SmallVector<mlir::Value> getOpOperandsSortedByUses(Operation* operation) {
   return map_to_vector(operandsAndUses, [](auto operandAndUse) { return operandAndUse.first; });
 }
 
-mlir::Value getBestOutputTensorFromOperandsOrAllocate(PatternRewriter& rewriter, Operation* operation) {
+bool hasLaterUserInBlock(mlir::Value value, Operation* operation) {
+  for (Operation* user : value.getUsers()) {
+    if (user->getBlock() != operation->getBlock())
+      return true;
+    if (operation->isBeforeInBlock(user))
+      return true;
+  }
+  return false;
+}
+
+mlir::Value getBestOutputTensorFromOperandsOrAllocate(RewriterBase& rewriter, Operation* operation) {
   assert("Only support operations with a single result" && operation->getNumResults() == 1);
   mlir::Value result = operation->getResult(0);
   auto resultType = result.getType();
   assert("Only support result ShapedType as result type" && isa<ShapedType>(resultType));
 
   SmallVector<mlir::Value> operands = getOpOperandsSortedByUses(operation);
-  auto validOperands =
+  auto validOperands = make_filter_range(operands, [operation, resultType](mlir::Value operand) {
-    make_filter_range(operands, [resultType](mlir::Value operand) { return operand.getType() == resultType; });
+    return operand.getType() == resultType && !hasLaterUserInBlock(operand, operation);
+  });
   auto bestOperand = validOperands.begin();
 
   if (bestOperand != validOperands.end())

src/PIM/Conversion/SpatialToPim/Common.hpp

@@ -2,16 +2,10 @@
 
 #include "mlir/Dialect/Tensor/IR/Tensor.h"
 
-#include "llvm/ADT/StringRef.h"
-
 #include "src/Accelerators/PIM/Common/PimCommon.hpp"
-#include "src/Accelerators/PIM/Dialect/Spatial/SpatialOps.hpp"
 
 namespace onnx_mlir {
 
-inline constexpr llvm::StringLiteral kChannelSourceCoreIdAttrName = "precomp_source_core_id";
-inline constexpr llvm::StringLiteral kChannelTargetCoreIdAttrName = "precomp_target_core_id";
-
 /**
  * \brief Get the offset of the ExtractSliceOp based on its static offsets and
  * its static tensor input.
@@ -30,17 +24,6 @@ size_t getShapedTypeSizeInBytes(mlir::ShapedType shapedType);
 
 mlir::IntegerAttr getTensorSizeInBytesAttr(mlir::Builder& builder, mlir::Value value);
 
-mlir::IntegerAttr getSpatialChannelSourceCoreIdAttr(mlir::Builder& builder, mlir::Value channel);
-
-mlir::IntegerAttr getSpatialChannelTargetCoreIdAttr(mlir::Builder& builder, mlir::Value channel);
-
-bool hasSpatialChannelSourceCoreIdAttr(mlir::Value channel);
-
-bool hasSpatialChannelTargetCoreIdAttr(mlir::Value channel);
-
-mlir::Value createPimReceiveFromSpatialChannel(
-  mlir::PatternRewriter& rewriter, mlir::Location loc, mlir::Value output, mlir::Value channel);
-
 template <class T>
 size_t rangeLength(const mlir::iterator_range<T> range) {
   return std::distance(range.begin(), range.end());
@@ -58,7 +41,7 @@ mlir::Operation* getEarliestUserWithinBlock(mlir::Value value);
 
 mlir::SmallVector<mlir::Value> getOpOperandsSortedByUses(mlir::Operation* operation);
 
-mlir::Value getBestOutputTensorFromOperandsOrAllocate(mlir::PatternRewriter& rewriter, mlir::Operation* operation);
+mlir::Value getBestOutputTensorFromOperandsOrAllocate(mlir::RewriterBase& rewriter, mlir::Operation* operation);
 
 inline mlir::tensor::EmptyOp
 createEmptyTensorFromShaped(mlir::IRRewriter& rewriter, mlir::Location loc, mlir::ShapedType shapedType) {

src/PIM/Conversion/SpatialToPim/ComputeLikeRegionUtils.cpp

@@ -0,0 +1,44 @@
+#include "src/Accelerators/PIM/Conversion/SpatialToPim/ComputeLikeRegionUtils.hpp"
+#include "src/Accelerators/PIM/Dialect/Spatial/SpatialOps.hpp"
+
+using namespace mlir;
+
+namespace onnx_mlir {
+
+std::optional<unsigned> getDirectComputeLikeInputIndex(Operation* owner, unsigned operandNumber) {
+  auto getInputIndex = [operandNumber](Operation* op, unsigned inputCount) -> std::optional<unsigned> {
+    if (inputCount == 0)
+      return std::nullopt;
+    unsigned inputBegin = op->getNumOperands() - inputCount;
+    if (operandNumber < inputBegin)
+      return std::nullopt;
+    return operandNumber - inputBegin;
+  };
+
+  if (auto compute = dyn_cast<spatial::SpatCompute>(owner))
+    return getInputIndex(owner, compute.getInputs().size());
+
+  if (auto computeBatch = dyn_cast<spatial::SpatComputeBatch>(owner))
+    return getInputIndex(owner, computeBatch.getInputs().size());
+
+  return std::nullopt;
+}
+
+void replaceAndEraseDirectComputeLikeInput(PatternRewriter& rewriter,
+                                           Operation* owner,
+                                           unsigned inputIndex,
+                                           Value replacement) {
+  Block& body = owner->getRegion(0).front();
+  BlockArgument bodyArgument = body.getArgument(inputIndex);
+
+  rewriter.startOpModification(owner);
+  bodyArgument.replaceAllUsesWith(replacement);
+  if (auto compute = dyn_cast<spatial::SpatCompute>(owner))
+    compute.getInputsMutable().erase(inputIndex);
+  else
+    cast<spatial::SpatComputeBatch>(owner).getInputsMutable().erase(inputIndex);
+  body.eraseArgument(inputIndex);
+  rewriter.finalizeOpModification(owner);
+}
+
+} // namespace onnx_mlir

src/PIM/Conversion/SpatialToPim/ComputeLikeRegionUtils.hpp

@@ -0,0 +1,17 @@
+#pragma once
+
+#include "mlir/IR/Operation.h"
+#include "mlir/IR/PatternMatch.h"
+
+#include <optional>
+
+namespace onnx_mlir {
+
+std::optional<unsigned> getDirectComputeLikeInputIndex(mlir::Operation* owner, unsigned operandNumber);
+
+void replaceAndEraseDirectComputeLikeInput(mlir::PatternRewriter& rewriter,
+                                           mlir::Operation* owner,
+                                           unsigned inputIndex,
+                                           mlir::Value replacement);
+
+} // namespace onnx_mlir

src/PIM/Conversion/SpatialToPim/CoreLoweringPatterns.cpp

@@ -0,0 +1,213 @@
+#include "mlir/Dialect/Arith/IR/Arith.h"
+#include "mlir/Dialect/Func/IR/FuncOps.h"
+#include "mlir/Dialect/Tensor/IR/Tensor.h"
+#include "mlir/Dialect/Tosa/IR/TosaOps.h"
+#include "mlir/IR/IRMapping.h"
+
+#include "Conversion/ONNXToSpatial/Common/Common.hpp"
+#include "src/Accelerators/PIM/Common/PimCommon.hpp"
+#include "src/Accelerators/PIM/Conversion/SpatialToPim/Common.hpp"
+#include "src/Accelerators/PIM/Conversion/SpatialToPim/CoreLoweringPatterns.hpp"
+#include "src/Accelerators/PIM/Dialect/Pim/PimOps.hpp"
+#include "src/Dialect/ONNX/ONNXOps.hpp"
+
+using namespace mlir;
+using namespace onnx_mlir::pim;
+
+namespace onnx_mlir {
+namespace {
+
+static bool isChannelUseChainOp(Operation* op) {
+  return isa<tensor::ExtractSliceOp,
+             tensor::CollapseShapeOp,
+             tensor::ExpandShapeOp,
+             tensor::CastOp,
+             tosa::ReshapeOp,
+             ONNXTransposeOp,
+             pim::PimTransposeOp>(op);
+}
+
+static void cloneMappedHelperOperands(Operation* op, IRMapping& mapping, IRRewriter& rewriter) {
+  for (Value operand : op->getOperands()) {
+    if (mapping.lookupOrNull(operand))
+      continue;
+
+    Operation* definingOp = operand.getDefiningOp();
+    if (!definingOp)
+      continue;
+
+    if (!isa<tensor::EmptyOp, arith::ConstantOp>(definingOp))
+      continue;
+
+    Operation* clonedOp = rewriter.clone(*definingOp, mapping);
+    for (auto [originalResult, newResult] : llvm::zip(definingOp->getResults(), clonedOp->getResults()))
+      mapping.map(originalResult, newResult);
+    rewriter.setInsertionPointAfter(clonedOp);
+  }
+}
+
+static int32_t translateSpatialCoreIdToPimCoreId(size_t spatialCoreId) { return static_cast<int32_t>(spatialCoreId); }
+
+static int32_t getPimCoreIdForComputeOp(spatial::SpatCompute computeOp, size_t& fallbackCoreId) {
+  if (auto spatialCoreIdAttr = computeOp->getAttrOfType<IntegerAttr>(onnx_mlir::kCoreIdAttrName))
+    return static_cast<int32_t>(spatialCoreIdAttr.getInt());
+  return static_cast<int32_t>(fallbackCoreId++);
+}
+
+static LogicalResult collectHelperComputeChain(spatial::SpatCompute computeOp,
+                                               SmallVectorImpl<Operation*>& helperChain,
+                                               bool requireReturnUse = true) {
+  if (computeOp.getInputs().size() != 1 || computeOp.getNumResults() != 1)
+    return failure();
+  if (requireReturnUse
+      && (!computeOp.getResult(0).hasOneUse() || !isa<func::ReturnOp>(*computeOp.getResult(0).getUsers().begin())))
+    return failure();
+
+  Block& block = computeOp.getBody().front();
+  if (block.getNumArguments() != 1)
+    return failure();
+
+  auto yieldOp = dyn_cast<spatial::SpatYieldOp>(block.getTerminator());
+  if (!yieldOp || yieldOp.getNumOperands() != 1)
+    return failure();
+
+  SmallVector<Operation*> reverseChain;
+  Value currentValue = yieldOp.getOperands().front();
+  Value blockArg = block.getArgument(0);
+
+  while (currentValue != blockArg) {
+    Operation* definingOp = currentValue.getDefiningOp();
+    if (!definingOp || definingOp->getBlock() != &block || !isChannelUseChainOp(definingOp))
+      return failure();
+    reverseChain.push_back(definingOp);
+    currentValue = definingOp->getOperand(0);
+  }
+
+  SmallPtrSet<Operation*, 8> chainSet(reverseChain.begin(), reverseChain.end());
+  for (Operation& op : llvm::make_early_inc_range(block.without_terminator()))
+    if (!chainSet.contains(&op) && !isa<tensor::EmptyOp, arith::ConstantOp>(op))
+      return failure();
+
+  helperChain.assign(reverseChain.rbegin(), reverseChain.rend());
+  return success();
+}
+
+static bool inlineInputlessHelperComputeForWeightLikeUsers(spatial::SpatCompute computeOp, IRRewriter& rewriter) {
+  if (!computeOp.getInputs().empty() || computeOp.getNumResults() != 1)
+    return false;
+  if (!llvm::all_of(computeOp.getResult(0).getUsers(), [](Operation* user) {
+        return isa<spatial::SpatCompute, spatial::SpatComputeBatch, pim::PimCoreOp, pim::PimCoreBatchOp>(user);
+      }))
+    return false;
+
+  Block& block = computeOp.getBody().front();
+  if (block.getNumArguments() != 0)
+    return false;
+
+  auto yieldOp = dyn_cast<spatial::SpatYieldOp>(block.getTerminator());
+  if (!yieldOp || yieldOp.getNumOperands() != 1)
+    return false;
+
+  rewriter.setInsertionPoint(computeOp);
+  IRMapping mapping;
+  for (Operation& op : block.without_terminator()) {
+    cloneMappedHelperOperands(&op, mapping, rewriter);
+    Operation* clonedOp = rewriter.clone(op, mapping);
+    for (auto [originalResult, newResult] : llvm::zip(op.getResults(), clonedOp->getResults()))
+      mapping.map(originalResult, newResult);
+    rewriter.setInsertionPointAfter(clonedOp);
+  }
+
+  Value replacement = mapping.lookupOrDefault(yieldOp.getOperand(0));
+  computeOp.getResult(0).replaceAllUsesWith(replacement);
+  return true;
+}
+
+} // namespace
+
+void markOpToRemove(CoreLoweringState& state, Operation* op) {
+  if (!llvm::is_contained(state.operationsToRemove, op))
+    state.operationsToRemove.push_back(op);
+}
+
+LogicalResult lowerComputeOp(spatial::SpatCompute computeOp, CoreLoweringState& state, IRRewriter& rewriter) {
+  Location loc = computeOp->getLoc();
+
+  if (inlineInputlessHelperComputeForWeightLikeUsers(computeOp, rewriter))
+    return success();
+
+  SmallVector<Operation*> helperChain;
+  if (succeeded(collectHelperComputeChain(computeOp, helperChain)))
+    return success();
+
+  auto& block = computeOp.getRegion().front();
+  auto yieldOp = cast<spatial::SpatYieldOp>(block.getTerminator());
+
+  for (auto [argIndex, blockArg] : llvm::enumerate(block.getArguments())) {
+    auto receiveOp = dyn_cast_or_null<spatial::SpatChannelReceiveOp>(computeOp.getInputs()[argIndex].getDefiningOp());
+    if (!receiveOp || blockArg.use_empty())
+      continue;
+
+    rewriter.setInsertionPoint(getEarliestUserWithinBlock(blockArg));
+    auto outputType = cast<ShapedType>(blockArg.getType());
+    auto outputBuffer = createEmptyTensorFromShaped(rewriter, receiveOp.getLoc(), outputType);
+    auto sizeAttr = getTensorSizeInBytesAttr(rewriter, blockArg);
+    auto sourceCoreIdAttr = rewriter.getI32IntegerAttr(translateSpatialCoreIdToPimCoreId(receiveOp.getSourceCoreId()));
+    Value received = PimReceiveOp::create(
+                       rewriter, receiveOp.getLoc(), outputBuffer.getType(), outputBuffer, sizeAttr, sourceCoreIdAttr)
+                       .getOutput();
+    blockArg.replaceAllUsesWith(received);
+    markOpToRemove(state, receiveOp);
+  }
+
+  if (computeOp.getNumResults() != yieldOp.getNumOperands())
+    llvm_unreachable("ComputeOp must have same number of results as yieldOp operands");
+
+  for (auto [result, yieldValue] : llvm::zip(computeOp.getResults(), yieldOp.getOperands())) {
+    if (result.use_empty())
+      continue;
+
+    ReturnPathState returnPathState {state.outputTensors, state.operationsToRemove};
+    ReturnPathLoweringResult returnPathResult =
+      lowerComputeResultReturnPath(computeOp, cast<OpResult>(result), yieldValue, returnPathState, rewriter);
+    if (returnPathResult == ReturnPathLoweringResult::Failure)
+      return failure();
+    if (returnPathResult == ReturnPathLoweringResult::Handled)
+      continue;
+
+    auto resultUses = result.getUses();
+    if (rangeLength(resultUses) == 1) {
+      OpOperand& resultUse = *resultUses.begin();
+      Operation* resultUser = resultUse.getOwner();
+      if (isa<spatial::SpatChannelSendOp>(resultUser))
+        continue;
+    }
+
+    return computeOp.emitOpError("has an unsupported remaining result use during Spatial-to-PIM lowering");
+  }
+
+  rewriter.setInsertionPoint(yieldOp);
+  rewriter.replaceOpWithNewOp<PimHaltOp>(yieldOp);
+
+  SmallVector<Value> computeWeights;
+  if (!computeOp.getWeights().empty())
+    computeWeights.append(computeOp.getWeights().begin(), computeOp.getWeights().end());
+  rewriter.setInsertionPointAfter(computeOp);
+  auto coreOp = PimCoreOp::create(rewriter,
+                                  loc,
+                                  ValueRange(computeWeights),
+                                  rewriter.getI32IntegerAttr(getPimCoreIdForComputeOp(computeOp, state.nextCoreId)));
+  auto& coreOpBlocks = coreOp.getBody().getBlocks();
+  for (auto [argIndex, blockArg] : llvm::enumerate(block.getArguments()))
+    if (!blockArg.use_empty())
+      blockArg.replaceAllUsesWith(computeOp.getInputs()[argIndex]);
+  block.eraseArguments(0, block.getNumArguments());
+  coreOpBlocks.splice(coreOpBlocks.begin(), computeOp.getBody().getBlocks());
+  Block* tempComputeBlock = new Block();
+  computeOp.getBody().push_back(tempComputeBlock);
+  rewriter.setInsertionPointToEnd(tempComputeBlock);
+  PimHaltOp::create(rewriter, computeOp.getLoc());
+  return success();
+}
+
+} // namespace onnx_mlir

src/PIM/Conversion/SpatialToPim/CoreLoweringPatterns.hpp

@@ -0,0 +1,21 @@
+#pragma once
+
+#include "mlir/IR/PatternMatch.h"
+
+#include "src/Accelerators/PIM/Conversion/SpatialToPim/ReturnPathNormalization.hpp"
+#include "src/Accelerators/PIM/Dialect/Spatial/SpatialOps.hpp"
+
+namespace onnx_mlir {
+
+struct CoreLoweringState {
+  size_t& nextCoreId;
+  llvm::SmallVectorImpl<OutputTensorFactory>& outputTensors;
+  llvm::SmallVectorImpl<mlir::Operation*>& operationsToRemove;
+};
+
+void markOpToRemove(CoreLoweringState& state, mlir::Operation* op);
+
+mlir::LogicalResult
+lowerComputeOp(spatial::SpatCompute computeOp, CoreLoweringState& state, mlir::IRRewriter& rewriter);
+
+} // namespace onnx_mlir

src/PIM/Conversion/SpatialToPim/GlobalTensorMaterialization.cpp

@@ -0,0 +1,390 @@
+#include "mlir/Dialect/Arith/IR/Arith.h"
+#include "mlir/Dialect/Bufferization/IR/Bufferization.h"
+#include "mlir/Dialect/Func/IR/FuncOps.h"
+#include "mlir/Dialect/MemRef/IR/MemRef.h"
+#include "mlir/Dialect/Tensor/IR/Tensor.h"
+#include "mlir/IR/BuiltinOps.h"
+#include "mlir/IR/BuiltinTypes.h"
+#include "mlir/IR/PatternMatch.h"
+#include "mlir/IR/SymbolTable.h"
+#include "mlir/IR/Value.h"
+
+#include "llvm/ADT/DenseMap.h"
+#include "llvm/ADT/STLExtras.h"
+#include "llvm/Support/Casting.h"
+#include "llvm/Support/LogicalResult.h"
+
+#include "Common/PimCommon.hpp"
+#include "src/Accelerators/PIM/Conversion/SpatialToPim/ComputeLikeRegionUtils.hpp"
+#include "src/Accelerators/PIM/Conversion/SpatialToPim/GlobalTensorMaterialization.hpp"
+#include "src/Accelerators/PIM/Dialect/Spatial/SpatialOps.hpp"
+
+using namespace mlir;
+
+namespace onnx_mlir {
+namespace {
+
+static std::string makeUniqueSymbolName(Operation* symbolTableOp, StringRef baseName) {
+  std::string name = baseName.str();
+  unsigned suffix = 0;
+  while (SymbolTable::lookupSymbolIn(symbolTableOp, name))
+    name = (baseName + "_" + Twine(suffix++)).str();
+  return name;
+}
+
+static memref::GlobalOp createPrivateMemrefGlobalWithUniqueName(PatternRewriter& rewriter,
+                                                                Location loc,
+                                                                ModuleOp moduleOp,
+                                                                StringRef baseName,
+                                                                MemRefType type,
+                                                                Attribute initialValue = {},
+                                                                UnitAttr constant = {}) {
+  std::string symbolName = makeUniqueSymbolName(moduleOp, baseName);
+  return memref::GlobalOp::create(rewriter,
+                                  loc,
+                                  rewriter.getStringAttr(symbolName),
+                                  rewriter.getStringAttr("private"),
+                                  TypeAttr::get(type),
+                                  initialValue,
+                                  constant,
+                                  IntegerAttr {});
+}
+
+// Sinks top-level tensor slices into compute regions so later lowering sees local runtime work.
+struct MoveExtractSliceIntoCompute final : OpRewritePattern<mlir::tensor::ExtractSliceOp> {
+  using OpRewritePattern::OpRewritePattern;
+
+  LogicalResult matchAndRewrite(mlir::tensor::ExtractSliceOp extractSliceOp, PatternRewriter& rewriter) const override {
+    if (!isa<func::FuncOp>(extractSliceOp->getParentOp()))
+      return failure();
+
+    for (auto& uses : extractSliceOp->getUses()) {
+      if (isa<spatial::SpatCompute>(uses.getOwner())) {
+        if (!getDirectComputeLikeInputIndex(uses.getOwner(), uses.getOperandNumber()))
+          return failure();
+      }
+      else if (isa_and_present<func::FuncOp>(uses.getOwner()->getParentOp())) {
+        return failure();
+      }
+    }
+
+    llvm::DenseMap<Operation*, Value> mapSpatToExtract;
+
+    for (auto& uses : llvm::make_early_inc_range(extractSliceOp->getUses())) {
+
+      if (auto spatCompute = dyn_cast<spatial::SpatCompute>(uses.getOwner())) {
+        auto inputIndex = getDirectComputeLikeInputIndex(spatCompute, uses.getOperandNumber());
+        if (!inputIndex)
+          return failure();
+        auto BBArgIndex = *inputIndex;
+        auto BBArgValue = spatCompute.getBody().front().getArgument(BBArgIndex);
+
+        if (BBArgValue.use_empty())
+          continue;
+
+        rewriter.setInsertionPoint(&spatCompute.getBody().front().front());
+        if (!mapSpatToExtract.contains(spatCompute.getOperation())) {
+          auto newExtractSlice = rewriter.clone(*extractSliceOp.getOperation());
+          mapSpatToExtract.insert({spatCompute.getOperation(), newExtractSlice->getResult(0)});
+        }
+
+        replaceAndEraseDirectComputeLikeInput(
+          rewriter, spatCompute.getOperation(), BBArgIndex, mapSpatToExtract[spatCompute.getOperation()]);
+      }
+      else if (auto spatComputeBatch = dyn_cast<spatial::SpatComputeBatch>(uses.getOwner())) {
+        auto inputIndex = getDirectComputeLikeInputIndex(spatComputeBatch, uses.getOperandNumber());
+        if (!inputIndex)
+          return failure();
+        auto BBArgIndex = *inputIndex;
+        auto BBArgValue = spatComputeBatch.getBody().front().getArgument(BBArgIndex);
+
+        if (BBArgValue.use_empty())
+          continue;
+
+        rewriter.setInsertionPoint(&spatComputeBatch.getBody().front().front());
+        if (!mapSpatToExtract.contains(spatComputeBatch.getOperation())) {
+          auto newExtractSlice = rewriter.clone(*extractSliceOp.getOperation());
+          mapSpatToExtract.insert({spatComputeBatch.getOperation(), newExtractSlice->getResult(0)});
+        }
+
+        replaceAndEraseDirectComputeLikeInput(
+          rewriter, spatComputeBatch.getOperation(), BBArgIndex, mapSpatToExtract[spatComputeBatch.getOperation()]);
+      }
+      else {
+        {
+          if (auto spatCompute = uses.getOwner()->getParentOfType<spatial::SpatCompute>()) {
+            rewriter.setInsertionPoint(&spatCompute.getBody().front().front());
+            if (!mapSpatToExtract.contains(spatCompute.getOperation())) {
+              auto newExtractSlice = rewriter.clone(*extractSliceOp.getOperation());
+              mapSpatToExtract.insert({spatCompute.getOperation(), newExtractSlice->getResult(0)});
+            }
+
+            rewriter.startOpModification(spatCompute.getOperation());
+            uses.set(mapSpatToExtract[spatCompute.getOperation()]);
+            rewriter.finalizeOpModification(spatCompute.getOperation());
+          }
+          else if (auto spatComputeBatch = uses.getOwner()->getParentOfType<spatial::SpatComputeBatch>()) {
+            rewriter.setInsertionPoint(&spatComputeBatch.getBody().front().front());
+            if (!mapSpatToExtract.contains(spatComputeBatch.getOperation())) {
+              auto newExtractSlice = rewriter.clone(*extractSliceOp.getOperation());
+              mapSpatToExtract.insert({spatComputeBatch.getOperation(), newExtractSlice->getResult(0)});
+            }
+
+            rewriter.startOpModification(spatComputeBatch.getOperation());
+            uses.set(mapSpatToExtract[spatComputeBatch.getOperation()]);
+            rewriter.finalizeOpModification(spatComputeBatch.getOperation());
+          }
+        }
+      }
+    }
+
+    rewriter.eraseOp(extractSliceOp);
+    return success();
+  }
+};
+
+// Turns runtime constants consumed by compute regions into private globals and local loads.
+struct ArithConstToGlobalMemoryPattern final : OpRewritePattern<mlir::arith::ConstantOp> {
+  using OpRewritePattern::OpRewritePattern;
+
+  LogicalResult matchAndRewrite(mlir::arith::ConstantOp constantOp, PatternRewriter& rewriter) const override {
+    Location loc = constantOp.getLoc();
+
+    if (hasWeightAlways(constantOp))
+      return failure();
+
+    if (!isa<func::FuncOp>(constantOp->getParentOp()))
+      return failure();
+
+    if (llvm::all_of(constantOp->getUsers(), [](Operation* op) {
+          if (isa<spatial::SpatCompute>(op))
+            return false;
+          if (isa<func::FuncOp>(op->getParentOp()))
+            return true;
+          return false;
+        }))
+      return failure();
+
+    rewriter.setInsertionPoint(constantOp->getParentOfType<func::FuncOp>());
+
+    auto constRankedTensorType = llvm::dyn_cast<mlir::RankedTensorType>(constantOp.getType());
+
+    if (constRankedTensorType) {
+      mlir::MemRefType memRefType =
+        mlir::MemRefType::get(constRankedTensorType.getShape(), constRankedTensorType.getElementType());
+      auto globalOp = createPrivateMemrefGlobalWithUniqueName(rewriter,
+                                                              loc,
+                                                              constantOp->getParentOfType<ModuleOp>(),
+                                                              "const",
+                                                              memRefType,
+                                                              constantOp.getValueAttr(),
+                                                              rewriter.getUnitAttr());
+      std::string argName = globalOp.getSymName().str();
+
+      llvm::DenseMap<Operation*, Value> mapSpatComputeToConst;
+
+      for (auto& constUses : llvm::make_early_inc_range(constantOp->getUses())) {
+        auto constUsers = constUses.getOwner();
+
+        if (auto spatCompute = llvm::dyn_cast<spatial::SpatCompute>(constUsers)) {
+          auto inputIndex = getDirectComputeLikeInputIndex(spatCompute, constUses.getOperandNumber());
+          if (!inputIndex)
+            return failure();
+          auto BBArgIndex = *inputIndex;
+          rewriter.setInsertionPoint(&spatCompute.getBody().front().front());
+          if (!mapSpatComputeToConst.contains(spatCompute.getOperation())) {
+            auto getGlobalOp = memref::GetGlobalOp::create(rewriter, loc, memRefType, argName);
+            auto toTensor = bufferization::ToTensorOp::create(
+              rewriter, loc, constRankedTensorType, getGlobalOp, rewriter.getUnitAttr(), rewriter.getUnitAttr());
+            mapSpatComputeToConst.insert({spatCompute.getOperation(), toTensor.getResult()});
+          }
+
+          replaceAndEraseDirectComputeLikeInput(
+            rewriter, spatCompute.getOperation(), BBArgIndex, mapSpatComputeToConst[spatCompute.getOperation()]);
+        }
+        else if (auto spatComputeBatch = llvm::dyn_cast<spatial::SpatComputeBatch>(constUsers)) {
+          auto inputIndex = getDirectComputeLikeInputIndex(spatComputeBatch, constUses.getOperandNumber());
+          if (!inputIndex)
+            return failure();
+          auto BBArgIndex = *inputIndex;
+          rewriter.setInsertionPoint(&spatComputeBatch.getBody().front().front());
+          if (!mapSpatComputeToConst.contains(spatComputeBatch.getOperation())) {
+            auto getGlobalOp = memref::GetGlobalOp::create(rewriter, loc, memRefType, argName);
+            auto toTensor = bufferization::ToTensorOp::create(
+              rewriter, loc, constRankedTensorType, getGlobalOp, rewriter.getUnitAttr(), rewriter.getUnitAttr());
+            mapSpatComputeToConst.insert({spatComputeBatch.getOperation(), toTensor.getResult()});
+          }
+
+          replaceAndEraseDirectComputeLikeInput(rewriter,
+                                                spatComputeBatch.getOperation(),
+                                                BBArgIndex,
+                                                mapSpatComputeToConst[spatComputeBatch.getOperation()]);
+        }
+        else {
+          {
+
+            if (auto spatCompute = constUses.getOwner()->getParentOfType<spatial::SpatCompute>()) {
+              rewriter.setInsertionPoint(&spatCompute.getBody().front().front());
+              if (!mapSpatComputeToConst.contains(spatCompute.getOperation())) {
+                auto getGlobalOp = memref::GetGlobalOp::create(rewriter, loc, memRefType, argName);
+                auto toTensor = bufferization::ToTensorOp::create(
+                  rewriter, loc, constRankedTensorType, getGlobalOp, rewriter.getUnitAttr(), rewriter.getUnitAttr());
+                mapSpatComputeToConst.insert({spatCompute.getOperation(), toTensor.getResult()});
+              }
+
+              rewriter.startOpModification(spatCompute.getOperation());
+              constUses.set(mapSpatComputeToConst[spatCompute.getOperation()]);
+              rewriter.finalizeOpModification(spatCompute.getOperation());
+            }
+            else if (auto spatComputeBatch = constUses.getOwner()->getParentOfType<spatial::SpatComputeBatch>()) {
+              rewriter.setInsertionPoint(&spatComputeBatch.getBody().front().front());
+              if (!mapSpatComputeToConst.contains(spatComputeBatch.getOperation())) {
+                auto getGlobalOp = memref::GetGlobalOp::create(rewriter, loc, memRefType, argName);
+                auto toTensor = bufferization::ToTensorOp::create(
+                  rewriter, loc, constRankedTensorType, getGlobalOp, rewriter.getUnitAttr(), rewriter.getUnitAttr());
+                mapSpatComputeToConst.insert({spatComputeBatch.getOperation(), toTensor.getResult()});
+              }
+
+              rewriter.startOpModification(spatComputeBatch.getOperation());
+              constUses.set(mapSpatComputeToConst[spatComputeBatch.getOperation()]);
+              rewriter.finalizeOpModification(spatComputeBatch.getOperation());
+            }
+          }
+        }
+      }
+    }
+    else if (constantOp.getType().isIntOrIndexOrFloat()) {
+      llvm::DenseMap<Operation*, Value> mapSpatComputeToConst;
+
+      for (auto& constUses : llvm::make_early_inc_range(constantOp->getUses())) {
+        auto constUsers = constUses.getOwner();
+
+        if (auto spatCompute = llvm::dyn_cast<spatial::SpatCompute>(constUsers)) {
+          auto inputIndex = getDirectComputeLikeInputIndex(spatCompute, constUses.getOperandNumber());
+          if (!inputIndex)
+            return failure();
+          auto BBArgIndex = *inputIndex;
+          rewriter.setInsertionPoint(&spatCompute.getBody().front().front());
+          auto newConst = rewriter.clone(*constantOp);
+
+          replaceAndEraseDirectComputeLikeInput(
+            rewriter, spatCompute.getOperation(), BBArgIndex, newConst->getResult(0));
+        }
+        else if (auto spatComputeBatch = llvm::dyn_cast<spatial::SpatComputeBatch>(constUsers)) {
+          auto inputIndex = getDirectComputeLikeInputIndex(spatComputeBatch, constUses.getOperandNumber());
+          if (!inputIndex)
+            return failure();
+          auto BBArgIndex = *inputIndex;
+          rewriter.setInsertionPoint(&spatComputeBatch.getBody().front().front());
+          auto newConst = rewriter.clone(*constantOp);
+
+          replaceAndEraseDirectComputeLikeInput(
+            rewriter, spatComputeBatch.getOperation(), BBArgIndex, newConst->getResult(0));
+        }
+        else if (auto parent = constUsers->getParentOfType<spatial::SpatCompute>()) {
+          if (!mapSpatComputeToConst.contains(parent)) {
+            rewriter.setInsertionPoint(&parent.getBody().front().front());
+            auto newConst = rewriter.clone(*constantOp);
+            mapSpatComputeToConst.insert({parent.getOperation(), newConst->getResult(0)});
+          }
+          constUses.set(mapSpatComputeToConst[parent.getOperation()]);
+        }
+        else {
+          auto batchParent = constUsers->getParentOfType<spatial::SpatComputeBatch>();
+          assert(batchParent && "Global Constant used direcly not within a compute");
+          if (!mapSpatComputeToConst.contains(batchParent.getOperation())) {
+            rewriter.setInsertionPoint(&batchParent.getBody().front().front());
+            auto newConst = rewriter.clone(*constantOp);
+            mapSpatComputeToConst.insert({batchParent.getOperation(), newConst->getResult(0)});
+          }
+          constUses.set(mapSpatComputeToConst[batchParent.getOperation()]);
+        }
+      }
+    }
+    if (constantOp->use_empty())
+      rewriter.eraseOp(constantOp);
+    return success();
+  }
+};
+
+// Materializes public function tensor inputs as globals so compute bodies can load them uniformly.
+struct FuncOpArgToGlobalMemoryPattern final : OpRewritePattern<mlir::func::FuncOp> {
+  using OpRewritePattern::OpRewritePattern;
+
+  LogicalResult matchAndRewrite(mlir::func::FuncOp funcOp, PatternRewriter& rewriter) const override {
+
+    if (funcOp.getArguments().empty())
+      return failure();
+
+    if (llvm::all_of(funcOp.getArguments(),
+                     [](mlir::BlockArgument blockArgument) { return blockArgument.use_empty(); }))
+      return failure();
+
+    Location loc = funcOp.getLoc();
+
+    for (auto [index, arg] : llvm::enumerate(funcOp.getArguments())) {
+      if (arg.getUses().empty())
+        continue;
+
+      rewriter.setInsertionPoint(funcOp.getOperation());
+
+      assert(isa<mlir::RankedTensorType>(arg.getType()));
+
+      auto argRankedTensorType = llvm::dyn_cast<mlir::RankedTensorType>(arg.getType());
+      mlir::MemRefType memRefType =
+        mlir::MemRefType::get(argRankedTensorType.getShape(), argRankedTensorType.getElementType());
+
+      std::string baseName = ("arg_" + Twine(index)).str();
+      auto globalOp = createPrivateMemrefGlobalWithUniqueName(
+        rewriter, loc, funcOp->getParentOfType<ModuleOp>(), baseName, memRefType);
+      std::string argName = globalOp.getSymName().str();
+
+      for (auto& argUses : llvm::make_early_inc_range(arg.getUses())) {
+        auto argUser = argUses.getOwner();
+        if (auto spatCompute = dyn_cast<spatial::SpatCompute>(argUser)) {
+          auto inputIndex = getDirectComputeLikeInputIndex(spatCompute, argUses.getOperandNumber());
+          if (!inputIndex)
+            return failure();
+          auto BBArgIndex = *inputIndex;
+          rewriter.setInsertionPoint(&spatCompute.getBody().front().front());
+          auto getGlobalOp = memref::GetGlobalOp::create(rewriter, loc, memRefType, argName);
+          auto toTensor = bufferization::ToTensorOp::create(
+            rewriter, loc, argRankedTensorType, getGlobalOp, rewriter.getUnitAttr(), rewriter.getUnitAttr());
+
+          replaceAndEraseDirectComputeLikeInput(rewriter, spatCompute.getOperation(), BBArgIndex, toTensor);
+        }
+        else if (auto spatComputeBatch = dyn_cast<spatial::SpatComputeBatch>(argUser)) {
+          auto inputIndex = getDirectComputeLikeInputIndex(spatComputeBatch, argUses.getOperandNumber());
+          if (!inputIndex)
+            return failure();
+          auto BBArgIndex = *inputIndex;
+          rewriter.setInsertionPoint(&spatComputeBatch.getBody().front().front());
+          auto getGlobalOp = memref::GetGlobalOp::create(rewriter, loc, memRefType, argName);
+          auto toTensor = bufferization::ToTensorOp::create(
+            rewriter, loc, argRankedTensorType, getGlobalOp, rewriter.getUnitAttr(), rewriter.getUnitAttr());
+
+          replaceAndEraseDirectComputeLikeInput(rewriter, spatComputeBatch.getOperation(), BBArgIndex, toTensor);
+        }
+        else {
+          rewriter.setInsertionPoint(argUser);
+          auto getGlobalOp = memref::GetGlobalOp::create(rewriter, loc, memRefType, argName);
+          auto toTensor = bufferization::ToTensorOp::create(
+            rewriter, loc, argRankedTensorType, getGlobalOp, rewriter.getUnitAttr(), rewriter.getUnitAttr());
+          rewriter.startOpModification(argUser);
+          argUses.set(toTensor);
+          rewriter.finalizeOpModification(argUser);
+        }
+      }
+    }
+
+    return success();
+  }
+};
+
+} // namespace
+void populateGlobalTensorMaterializationPatterns(RewritePatternSet& patterns) {
+  patterns.add<MoveExtractSliceIntoCompute, FuncOpArgToGlobalMemoryPattern, ArithConstToGlobalMemoryPattern>(
+    patterns.getContext());
+}
+
+} // namespace onnx_mlir

src/PIM/Conversion/SpatialToPim/GlobalTensorMaterialization.hpp

@@ -0,0 +1,9 @@
+#pragma once
+
+#include "mlir/IR/PatternMatch.h"
+
+namespace onnx_mlir {
+
+void populateGlobalTensorMaterializationPatterns(mlir::RewritePatternSet& patterns);
+
+}

src/PIM/Conversion/SpatialToPim/PhaseVerification.cpp

@@ -0,0 +1,20 @@
+#include "src/Accelerators/PIM/Conversion/SpatialToPim/PhaseVerification.hpp"
+#include "src/Accelerators/PIM/Dialect/Spatial/SpatialOps.hpp"
+
+using namespace mlir;
+
+namespace onnx_mlir {
+
+LogicalResult verifySpatialToPimBoundary(ModuleOp moduleOp) {
+  bool hasFailure = false;
+  moduleOp.walk([&](Operation* op) {
+    if (op->getDialect()->getNamespace() != "spat")
+      return;
+
+    op->emitError("illegal Spatial operation remains after Spatial-to-PIM lowering");
+    hasFailure = true;
+  });
+  return success(!hasFailure);
+}
+
+} // namespace onnx_mlir

src/PIM/Conversion/SpatialToPim/PhaseVerification.hpp

@@ -0,0 +1,9 @@
+#pragma once
+
+#include "mlir/IR/BuiltinOps.h"
+
+namespace onnx_mlir {
+
+mlir::LogicalResult verifySpatialToPimBoundary(mlir::ModuleOp moduleOp);
+
+} // namespace onnx_mlir

src/PIM/Conversion/SpatialToPim/ReturnPathNormalization.cpp

@@ -0,0 +1,587 @@
+#include "mlir/Dialect/Arith/IR/Arith.h"
+#include "mlir/Dialect/Bufferization/IR/Bufferization.h"
+#include "mlir/Dialect/MemRef/IR/MemRef.h"
+#include "mlir/Dialect/Tensor/IR/Tensor.h"
+#include "mlir/Dialect/Tosa/IR/TosaOps.h"
+#include "mlir/IR/BuiltinOps.h"
+#include "mlir/IR/IRMapping.h"
+#include "mlir/IR/SymbolTable.h"
+
+#include "Conversion/ONNXToSpatial/Common/Common.hpp"
+#include "src/Accelerators/PIM/Conversion/SpatialToPim/Common.hpp"
+#include "src/Accelerators/PIM/Conversion/SpatialToPim/ReturnPathNormalization.hpp"
+#include "src/Accelerators/PIM/Dialect/Pim/PimOps.hpp"
+#include "src/Dialect/ONNX/ONNXOps.hpp"
+
+using namespace mlir;
+using namespace onnx_mlir::pim;
+
+namespace onnx_mlir {
+namespace {
+
+struct ReturnUseInfo {
+  size_t returnIndex;
+  SmallVector<Operation*> helperChain;
+};
+
+struct ConcatReturnUseInfo {
+  size_t returnIndex;
+  SmallVector<int64_t> sliceOffsets;
+  SmallVector<int64_t> concatShape;
+  SmallVector<Operation*> concatChain;
+  SmallVector<Operation*> helperChain;
+};
+
+static bool isReturnHelperChainOp(Operation* op) {
+  return isa<tensor::ExtractSliceOp,
+             tensor::CollapseShapeOp,
+             tensor::ExpandShapeOp,
+             tensor::CastOp,
+             tosa::ReshapeOp,
+             ONNXTransposeOp,
+             pim::PimTransposeOp>(op);
+}
+
+static void markOpToRemove(ReturnPathState& state, Operation* op) {
+  if (!llvm::is_contained(state.operationsToRemove, op))
+    state.operationsToRemove.push_back(op);
+}
+
+static std::string makeUniqueSymbolName(Operation* symbolTableOp, StringRef baseName) {
+  std::string name = baseName.str();
+  unsigned suffix = 0;
+  while (SymbolTable::lookupSymbolIn(symbolTableOp, name))
+    name = (baseName + "_" + Twine(suffix++)).str();
+  return name;
+}
+
+static int64_t computeFlatElementIndex(ArrayRef<int64_t> indices, ArrayRef<int64_t> shape) {
+  int64_t flatIndex = 0;
+  for (size_t i = 0; i < shape.size(); ++i) {
+    flatIndex *= shape[i];
+    flatIndex += indices[i];
+  }
+  return flatIndex;
+}
+
+static SmallVector<int64_t> expandFlatElementIndex(int64_t flatIndex, ArrayRef<int64_t> shape) {
+  SmallVector<int64_t> indices(shape.size(), 0);
+  for (int64_t dim = static_cast<int64_t>(shape.size()) - 1; dim >= 0; --dim) {
+    indices[dim] = flatIndex % shape[dim];
+    flatIndex /= shape[dim];
+  }
+  return indices;
+}
+
+static LogicalResult collectHelperComputeChain(spatial::SpatCompute computeOp,
+                                               SmallVectorImpl<Operation*>& helperChain) {
+  if (computeOp.getInputs().size() != 1 || computeOp.getNumResults() != 1)
+    return failure();
+  if (!computeOp.getResult(0).hasOneUse() || !isa<func::ReturnOp>(*computeOp.getResult(0).getUsers().begin()))
+    return failure();
+
+  Block& block = computeOp.getBody().front();
+  if (block.getNumArguments() != 1)
+    return failure();
+
+  auto yieldOp = dyn_cast<spatial::SpatYieldOp>(block.getTerminator());
+  if (!yieldOp || yieldOp.getNumOperands() != 1)
+    return failure();
+
+  SmallVector<Operation*> reverseChain;
+  Value currentValue = yieldOp.getOperands().front();
+  Value blockArg = block.getArgument(0);
+
+  while (currentValue != blockArg) {
+    Operation* definingOp = currentValue.getDefiningOp();
+    if (!definingOp || definingOp->getBlock() != &block || !isReturnHelperChainOp(definingOp))
+      return failure();
+    reverseChain.push_back(definingOp);
+    currentValue = definingOp->getOperand(0);
+  }
+
+  SmallPtrSet<Operation*, 8> chainSet(reverseChain.begin(), reverseChain.end());
+  for (Operation& op : llvm::make_early_inc_range(block.without_terminator()))
+    if (!chainSet.contains(&op) && !isa<tensor::EmptyOp, arith::ConstantOp>(op))
+      return failure();
+
+  helperChain.assign(reverseChain.rbegin(), reverseChain.rend());
+  return success();
+}
+
+static std::optional<ReturnUseInfo> analyzeReturnUse(Value value) {
+  auto uses = value.getUses();
+  if (rangeLength(uses) != 1)
+    return std::nullopt;
+
+  SmallVector<Operation*> helperChain;
+  Value currentValue = value;
+  Operation* currentUser = uses.begin()->getOwner();
+
+  while (isReturnHelperChainOp(currentUser)) {
+    helperChain.push_back(currentUser);
+    auto currentUses = currentUser->getResult(0).getUses();
+    if (rangeLength(currentUses) != 1)
+      return std::nullopt;
+    currentValue = currentUser->getResult(0);
+    currentUser = currentUses.begin()->getOwner();
+  }
+
+  if (!isa<func::ReturnOp>(currentUser))
+    return std::nullopt;
+
+  return ReturnUseInfo {
+    currentValue.getUses().begin()->getOperandNumber(),
+    std::move(helperChain),
+  };
+}
+
+static std::optional<ConcatReturnUseInfo> analyzeConcatReturnUse(Value value) {
+  auto getConcatResult = [](Operation* op) -> Value {
+    if (auto tensorConcat = dyn_cast<tensor::ConcatOp>(op))
+      return tensorConcat.getResult();
+    if (auto spatialConcat = dyn_cast<spatial::SpatConcatOp>(op))
+      return spatialConcat.getOutput();
+    if (auto pimConcat = dyn_cast<pim::PimConcatOp>(op))
+      return pimConcat.getOutput();
+    return {};
+  };
+  auto getConcatAxis = [](Operation* op) -> std::optional<int64_t> {
+    if (auto tensorConcat = dyn_cast<tensor::ConcatOp>(op))
+      return tensorConcat.getDim();
+    if (auto spatialConcat = dyn_cast<spatial::SpatConcatOp>(op))
+      return spatialConcat.getAxis();
+    if (auto pimConcat = dyn_cast<pim::PimConcatOp>(op))
+      return pimConcat.getAxis();
+    return std::nullopt;
+  };
+  auto getConcatOperands = [](Operation* op) -> OperandRange {
+    if (auto tensorConcat = dyn_cast<tensor::ConcatOp>(op))
+      return tensorConcat.getOperands();
+    if (auto spatialConcat = dyn_cast<spatial::SpatConcatOp>(op))
+      return spatialConcat.getInputs();
+    return cast<pim::PimConcatOp>(op).getInputs();
+  };
+
+  auto uses = value.getUses();
+  if (rangeLength(uses) != 1
+      || !isa<tensor::ConcatOp, spatial::SpatConcatOp, pim::PimConcatOp>(uses.begin()->getOwner()))
+    return std::nullopt;
+
+  auto valueType = dyn_cast<ShapedType>(value.getType());
+  if (!valueType || !valueType.hasStaticShape())
+    return std::nullopt;
+
+  SmallVector<int64_t> sliceOffsets(valueType.getRank(), 0);
+  SmallVector<int64_t> concatShape(valueType.getShape().begin(), valueType.getShape().end());
+  SmallVector<Operation*> concatChain;
+  Value currentValue = value;
+  Operation* currentUser = uses.begin()->getOwner();
+
+  while (isa<tensor::ConcatOp, spatial::SpatConcatOp, pim::PimConcatOp>(currentUser)) {
+    concatChain.push_back(currentUser);
+    size_t operandIndex = currentValue.getUses().begin()->getOperandNumber();
+    int64_t axis = *getConcatAxis(currentUser);
+    for (Value operand : getConcatOperands(currentUser).take_front(operandIndex))
+      sliceOffsets[axis] += cast<ShapedType>(operand.getType()).getShape()[axis];
+
+    Value concatResult = getConcatResult(currentUser);
+    auto concatType = dyn_cast<ShapedType>(concatResult.getType());
+    if (!concatType || !concatType.hasStaticShape())
+      return std::nullopt;
+    concatShape.assign(concatType.getShape().begin(), concatType.getShape().end());
+
+    currentValue = concatResult;
+    auto currentUses = currentValue.getUses();
+    if (rangeLength(currentUses) != 1)
+      return std::nullopt;
+    currentUser = currentUses.begin()->getOwner();
+  }
+
+  SmallVector<Operation*> helperChain;
+  if (auto helperCompute = dyn_cast<spatial::SpatCompute>(currentUser)) {
+    if (helperCompute.getInputs().size() != 1 || helperCompute.getInputs().front() != currentValue)
+      return std::nullopt;
+
+    if (failed(collectHelperComputeChain(helperCompute, helperChain)))
+      return std::nullopt;
+
+    currentValue = helperCompute.getResult(0);
+    auto currentUses = currentValue.getUses();
+    if (rangeLength(currentUses) != 1)
+      return std::nullopt;
+    currentUser = currentUses.begin()->getOwner();
+  }
+
+  while (isReturnHelperChainOp(currentUser)) {
+    helperChain.push_back(currentUser);
+    auto currentUses = currentUser->getResult(0).getUses();
+    if (rangeLength(currentUses) != 1)
+      return std::nullopt;
+    currentValue = currentUser->getResult(0);
+    currentUser = currentUses.begin()->getOwner();
+  }
+
+  if (!isa<func::ReturnOp>(currentUser))
+    return std::nullopt;
+
+  return ConcatReturnUseInfo {
+    currentValue.getUses().begin()->getOperandNumber(),
+    std::move(sliceOffsets),
+    std::move(concatShape),
+    std::move(concatChain),
+    std::move(helperChain),
+  };
+}
+
+static LogicalResult mapIndicesThroughHelperChain(ArrayRef<int64_t> sourceIndices,
+                                                  ArrayRef<int64_t> sourceShape,
+                                                  ArrayRef<Operation*> helperChain,
+                                                  SmallVectorImpl<int64_t>& mappedIndices) {
+  SmallVector<int64_t> currentIndices(sourceIndices.begin(), sourceIndices.end());
+  SmallVector<int64_t> currentShape(sourceShape.begin(), sourceShape.end());
+
+  auto reshapeToResultShape = [&](Operation* op) -> LogicalResult {
+    auto resultType = dyn_cast<ShapedType>(op->getResult(0).getType());
+    if (!resultType || !resultType.hasStaticShape())
+      return failure();
+    int64_t flatIndex = computeFlatElementIndex(currentIndices, currentShape);
+    currentShape.assign(resultType.getShape().begin(), resultType.getShape().end());
+    currentIndices = expandFlatElementIndex(flatIndex, currentShape);
+    return success();
+  };
+
+  for (Operation* op : helperChain) {
+    if (auto extractSliceOp = dyn_cast<tensor::ExtractSliceOp>(op)) {
+      auto hasStaticValues = [](ArrayRef<int64_t> values) {
+        return llvm::all_of(values, [](int64_t value) { return !ShapedType::isDynamic(value); });
+      };
+      if (!hasStaticValues(extractSliceOp.getStaticOffsets()) || !hasStaticValues(extractSliceOp.getStaticSizes())
+          || !hasStaticValues(extractSliceOp.getStaticStrides()))
+        return failure();
+
+      SmallVector<int64_t> nextIndices;
+      nextIndices.reserve(currentIndices.size());
+      for (auto [index, offset, size, stride] : llvm::zip_equal(currentIndices,
+                                                                extractSliceOp.getStaticOffsets(),
+                                                                extractSliceOp.getStaticSizes(),
+                                                                extractSliceOp.getStaticStrides())) {
+        if (stride != 1 || index < offset || index >= offset + size)
+          return failure();
+        nextIndices.push_back(index - offset);
+      }
+
+      auto resultType = dyn_cast<ShapedType>(extractSliceOp.getResult().getType());
+      if (!resultType || !resultType.hasStaticShape())
+        return failure();
+      currentIndices = std::move(nextIndices);
+      currentShape.assign(resultType.getShape().begin(), resultType.getShape().end());
+      continue;
+    }
+
+    if (auto transposeOp = dyn_cast<ONNXTransposeOp>(op)) {
+      SmallVector<int64_t> nextIndices(currentIndices.size());
+      SmallVector<int64_t> nextShape(currentShape.size());
+      for (auto [destIndex, attr] : llvm::enumerate(transposeOp.getPermAttr().getAsRange<IntegerAttr>())) {
+        int64_t sourceIndex = attr.getInt();
+        nextIndices[destIndex] = currentIndices[sourceIndex];
+        nextShape[destIndex] = currentShape[sourceIndex];
+      }
+      currentIndices = std::move(nextIndices);
+      currentShape = std::move(nextShape);
+      continue;
+    }
+
+    if (auto transposeOp = dyn_cast<pim::PimTransposeOp>(op)) {
+      SmallVector<int64_t> nextIndices(currentIndices.size());
+      SmallVector<int64_t> nextShape(currentShape.size());
+      for (auto [destIndex, attr] : llvm::enumerate(transposeOp.getPermutation().getAsRange<IntegerAttr>())) {
+        int64_t sourceIndex = attr.getInt();
+        nextIndices[destIndex] = currentIndices[sourceIndex];
+        nextShape[destIndex] = currentShape[sourceIndex];
+      }
+      currentIndices = std::move(nextIndices);
+      currentShape = std::move(nextShape);
+      continue;
+    }
+
+    if (isa<tensor::CastOp, tosa::ReshapeOp, tensor::CollapseShapeOp, tensor::ExpandShapeOp>(op)) {
+      if (failed(reshapeToResultShape(op)))
+        return failure();
+      continue;
+    }
+
+    return failure();
+  }
+
+  mappedIndices.assign(currentIndices.begin(), currentIndices.end());
+  return success();
+}
+
+static void cloneMappedHelperOperands(Operation* op, IRMapping& mapping, IRRewriter& rewriter) {
+  for (Value operand : op->getOperands()) {
+    if (mapping.lookupOrNull(operand))
+      continue;
+
+    Operation* definingOp = operand.getDefiningOp();
+    if (!definingOp)
+      continue;
+
+    if (!isa<tensor::EmptyOp, arith::ConstantOp>(definingOp))
+      continue;
+
+    Operation* clonedOp = rewriter.clone(*definingOp, mapping);
+    for (auto [originalResult, newResult] : llvm::zip(definingOp->getResults(), clonedOp->getResults()))
+      mapping.map(originalResult, newResult);
+    rewriter.setInsertionPointAfter(clonedOp);
+  }
+}
+
+static void
+cloneHelperChain(Value sourceValue, ArrayRef<Operation*> helperChain, IRRewriter& rewriter, Value& clonedValue) {
+  IRMapping mapping;
+  mapping.map(sourceValue, sourceValue);
+  clonedValue = sourceValue;
+
+  rewriter.setInsertionPointAfterValue(sourceValue);
+  for (Operation* op : helperChain) {
+    cloneMappedHelperOperands(op, mapping, rewriter);
+    Operation* clonedOp = rewriter.clone(*op, mapping);
+    for (auto [originalResult, newResult] : llvm::zip(op->getResults(), clonedOp->getResults()))
+      mapping.map(originalResult, newResult);
+    clonedValue = clonedOp->getResult(0);
+    rewriter.setInsertionPointAfter(clonedOp);
+  }
+}
+
+static Value emitHostCopy(IRRewriter& rewriter,
+                          Location loc,
+                          Value outputTensor,
+                          Value sourceValue,
+                          int32_t hostTargetOffset,
+                          int32_t deviceSourceOffset,
+                          int32_t sizeInBytes) {
+  return PimMemCopyDevToHostOp::create(rewriter,
+                                       loc,
+                                       outputTensor.getType(),
+                                       outputTensor,
+                                       sourceValue,
+                                       rewriter.getI32IntegerAttr(hostTargetOffset),
+                                       rewriter.getI32IntegerAttr(deviceSourceOffset),
+                                       rewriter.getI32IntegerAttr(sizeInBytes))
+    .getOutput();
+}
+
+} // namespace
+
+void addReturnOutputBuffers(func::ReturnOp returnOp,
+                            IRRewriter& rewriter,
+                            SmallVectorImpl<OutputTensorFactory>& outputTensors) {
+  outputTensors.reserve(returnOp->getNumOperands());
+  for (auto [index, returnValue] : llvm::enumerate(returnOp->getOperands())) {
+    Value currentReturnValue = returnValue;
+    Operation* returnValueDefiningOp = currentReturnValue.getDefiningOp();
+    if (returnValueDefiningOp->hasTrait<OpTrait::ConstantLike>()) {
+      assert(!hasWeightAlways(returnValueDefiningOp));
+      outputTensors.push_back(
+        [currentReturnValue](IRRewriter& rewriter, Location loc) -> Value { return currentReturnValue; });
+    }
+    else {
+      auto outRankedTensorType = llvm::dyn_cast<RankedTensorType>(currentReturnValue.getType());
+      auto memRefType = MemRefType::get(outRankedTensorType.getShape(), outRankedTensorType.getElementType());
+
+      std::string outputBaseName = ("output_" + Twine(index)).str();
+      std::string outputName = makeUniqueSymbolName(returnOp->getParentOfType<ModuleOp>(), outputBaseName);
+      rewriter.setInsertionPoint(returnOp.getParentOp());
+      memref::GlobalOp::create(rewriter,
+                               returnOp.getLoc(),
+                               rewriter.getStringAttr(outputName),
+                               rewriter.getStringAttr("private"),
+                               TypeAttr::get(memRefType),
+                               {},
+                               {},
+                               {});
+      outputTensors.push_back([memRefType, outputName, outRankedTensorType](IRRewriter& rewriter, Location loc) {
+        auto getGlobalOp = memref::GetGlobalOp::create(rewriter, loc, memRefType, outputName);
+        auto toTensor = bufferization::ToTensorOp::create(
+          rewriter, loc, outRankedTensorType, getGlobalOp, rewriter.getUnitAttr(), rewriter.getUnitAttr());
+        return toTensor.getResult();
+      });
+    }
+  }
+}
+
+ReturnPathLoweringResult lowerComputeResultReturnPath(
+  spatial::SpatCompute computeOp, OpResult result, Value yieldValue, ReturnPathState& state, IRRewriter& rewriter) {
+  Location loc = computeOp->getLoc();
+  auto yieldType = cast<TensorType>(yieldValue.getType());
+
+  if (auto returnUse = analyzeReturnUse(result)) {
+    Value storedValue = yieldValue;
+    cloneHelperChain(yieldValue, returnUse->helperChain, rewriter, storedValue);
+    for (Operation* op : returnUse->helperChain)
+      markOpToRemove(state, op);
+
+    auto storedType = cast<ShapedType>(storedValue.getType());
+    size_t elementSize = storedType.getElementTypeBitWidth() / 8;
+    if (auto storedOp = storedValue.getDefiningOp())
+      rewriter.setInsertionPointAfter(storedOp);
+    Value outputTensor = state.outputTensors[returnUse->returnIndex](rewriter, loc);
+    emitHostCopy(
+      rewriter, loc, outputTensor, storedValue, 0, 0, static_cast<int32_t>(storedType.getNumElements() * elementSize));
+    return ReturnPathLoweringResult::Handled;
+  }
+
+  auto resultUses = result.getUses();
+  if (rangeLength(resultUses) == 1) {
+    OpOperand& resultUse = *resultUses.begin();
+    Operation* resultUser = resultUse.getOwner();
+
+    if (isa<func::ReturnOp>(resultUser)) {
+      size_t resultIndexInReturn = resultUse.getOperandNumber();
+      size_t elementSize = yieldType.getElementType().getIntOrFloatBitWidth() / 8;
+      rewriter.setInsertionPointAfterValue(yieldValue);
+      Value outputTensor = state.outputTensors[resultIndexInReturn](rewriter, loc);
+      emitHostCopy(
+        rewriter, loc, outputTensor, yieldValue, 0, 0, static_cast<int32_t>(yieldType.getNumElements() * elementSize));
+      return ReturnPathLoweringResult::Handled;
+    }
+  }
+
+  if (auto concatReturnUse = analyzeConcatReturnUse(result)) {
+    size_t elementSize = yieldType.getElementTypeBitWidth() / 8;
+    for (Operation* concatOp : concatReturnUse->concatChain)
+      markOpToRemove(state, concatOp);
+
+    if (concatReturnUse->helperChain.empty()) {
+      rewriter.setInsertionPointAfterValue(yieldValue);
+      Value outputTensor = state.outputTensors[concatReturnUse->returnIndex](rewriter, loc);
+      auto outputType = cast<ShapedType>(outputTensor.getType());
+      int64_t flatOffset = computeFlatElementIndex(concatReturnUse->sliceOffsets, outputType.getShape());
+      emitHostCopy(rewriter,
+                   loc,
+                   outputTensor,
+                   yieldValue,
+                   static_cast<int32_t>(flatOffset * elementSize),
+                   0,
+                   static_cast<int32_t>(yieldType.getNumElements() * elementSize));
+      return ReturnPathLoweringResult::Handled;
+    }
+
+    auto storedType = dyn_cast<RankedTensorType>(yieldValue.getType());
+    if (!storedType) {
+      computeOp.emitOpError("has an unsupported non-ranked concat-return helper yield during Spatial-to-PIM lowering");
+      return ReturnPathLoweringResult::Failure;
+    }
+    rewriter.setInsertionPointAfterValue(yieldValue);
+    Value outputTensor = state.outputTensors[concatReturnUse->returnIndex](rewriter, loc);
+    auto outputType = cast<ShapedType>(outputTensor.getType());
+    for (int64_t linearIndex = 0; linearIndex < storedType.getNumElements(); ++linearIndex) {
+      SmallVector<int64_t> sourceIndices = expandFlatElementIndex(linearIndex, storedType.getShape());
+      for (auto [dim, idx] : llvm::enumerate(sourceIndices))
+        sourceIndices[dim] = concatReturnUse->sliceOffsets[dim] + idx;
+
+      SmallVector<int64_t> destinationIndices;
+      if (failed(mapIndicesThroughHelperChain(
+            sourceIndices, concatReturnUse->concatShape, concatReturnUse->helperChain, destinationIndices))) {
+        computeOp.emitOpError("has an unsupported concat-return helper chain during Spatial-to-PIM lowering");
+        return ReturnPathLoweringResult::Failure;
+      }
+
+      SmallVector<OpFoldResult> extractOffsets;
+      SmallVector<OpFoldResult> extractSizes;
+      SmallVector<OpFoldResult> extractStrides;
+      extractOffsets.reserve(storedType.getRank());
+      extractSizes.reserve(storedType.getRank());
+      extractStrides.reserve(storedType.getRank());
+      for (int64_t idx : expandFlatElementIndex(linearIndex, storedType.getShape())) {
+        extractOffsets.push_back(rewriter.getIndexAttr(idx));
+        extractSizes.push_back(rewriter.getIndexAttr(1));
+        extractStrides.push_back(rewriter.getIndexAttr(1));
+      }
+
+      auto scalarTensorType =
+        RankedTensorType::get(SmallVector<int64_t>(storedType.getRank(), 1), storedType.getElementType());
+      auto elementSlice = tensor::ExtractSliceOp::create(
+        rewriter, loc, scalarTensorType, yieldValue, extractOffsets, extractSizes, extractStrides);
+      rewriter.setInsertionPointAfter(elementSlice);
+
+      int64_t destinationFlatOffset = computeFlatElementIndex(destinationIndices, outputType.getShape());
+      outputTensor = emitHostCopy(rewriter,
+                                  loc,
+                                  outputTensor,
+                                  elementSlice.getResult(),
+                                  static_cast<int32_t>(destinationFlatOffset * elementSize),
+                                  0,
+                                  static_cast<int32_t>(elementSize));
+    }
+    return ReturnPathLoweringResult::Handled;
+  }
+
+  return ReturnPathLoweringResult::NotReturnPath;
+}
+
+void replaceReturnWithOutputBuffers(func::ReturnOp returnOp, IRRewriter& rewriter, ReturnPathState& state) {
+  auto markOwnedReturnChain = [&](Operation* op, auto&& markOwnedReturnChain) -> void {
+    if (!op)
+      return;
+
+    bool isExclusivelyOwnedByReturnChain = op->use_empty();
+    if (!isExclusivelyOwnedByReturnChain && op->hasOneUse()) {
+      Operation* onlyUser = *op->getUsers().begin();
+      isExclusivelyOwnedByReturnChain =
+        isa<func::ReturnOp, tensor::ConcatOp, spatial::SpatConcatOp, pim::PimConcatOp, spatial::SpatCompute>(onlyUser)
+        || isReturnHelperChainOp(onlyUser);
+    }
+    if (!isExclusivelyOwnedByReturnChain)
+      return;
+
+    if (isReturnHelperChainOp(op)) {
+      Value source = op->getOperand(0);
+      markOpToRemove(state, op);
+      markOwnedReturnChain(source.getDefiningOp(), markOwnedReturnChain);
+      return;
+    }
+
+    if (auto computeOp = dyn_cast<spatial::SpatCompute>(op)) {
+      markOpToRemove(state, computeOp);
+      if (!computeOp.getInputs().empty())
+        for (Value input : computeOp.getInputs())
+          markOwnedReturnChain(input.getDefiningOp(), markOwnedReturnChain);
+      return;
+    }
+
+    if (auto concatOp = dyn_cast<tensor::ConcatOp>(op)) {
+      markOpToRemove(state, concatOp);
+      for (Value operand : concatOp.getOperands())
+        markOwnedReturnChain(operand.getDefiningOp(), markOwnedReturnChain);
+      return;
+    }
+
+    if (auto concatOp = dyn_cast<spatial::SpatConcatOp>(op)) {
+      markOpToRemove(state, concatOp);
+      for (Value operand : concatOp.getInputs())
+        markOwnedReturnChain(operand.getDefiningOp(), markOwnedReturnChain);
+      return;
+    }
+
+    if (auto concatOp = dyn_cast<pim::PimConcatOp>(op)) {
+      markOpToRemove(state, concatOp);
+      for (Value operand : concatOp.getInputs())
+        markOwnedReturnChain(operand.getDefiningOp(), markOwnedReturnChain);
+    }
+  };
+
+  SmallVector<Value> originalOperands(returnOp.getOperands().begin(), returnOp.getOperands().end());
+  auto loc = returnOp.getLoc();
+  for (auto it : llvm::enumerate(originalOperands)) {
+    size_t orderWithinReturn = it.index();
+    Operation* returnOperand = it.value().getDefiningOp();
+    rewriter.setInsertionPoint(returnOp);
+    Value outputTensor = state.outputTensors[orderWithinReturn](rewriter, loc);
+    rewriter.modifyOpInPlace(returnOp, [&] { returnOp.setOperand(orderWithinReturn, outputTensor); });
+    markOwnedReturnChain(returnOperand, markOwnedReturnChain);
+  }
+}
+
+} // namespace onnx_mlir

src/PIM/Conversion/SpatialToPim/ReturnPathNormalization.hpp

@@ -0,0 +1,37 @@
+#pragma once
+
+#include "mlir/Dialect/Func/IR/FuncOps.h"
+#include "mlir/IR/PatternMatch.h"
+
+#include <functional>
+
+#include "src/Accelerators/PIM/Dialect/Spatial/SpatialOps.hpp"
+
+namespace onnx_mlir {
+
+using OutputTensorFactory = std::function<mlir::Value(mlir::IRRewriter& rewriter, mlir::Location loc)>;
+
+struct ReturnPathState {
+  llvm::SmallVectorImpl<OutputTensorFactory>& outputTensors;
+  llvm::SmallVectorImpl<mlir::Operation*>& operationsToRemove;
+};
+
+enum class ReturnPathLoweringResult {
+  Handled,
+  NotReturnPath,
+  Failure
+};
+
+void addReturnOutputBuffers(mlir::func::ReturnOp returnOp,
+                            mlir::IRRewriter& rewriter,
+                            llvm::SmallVectorImpl<OutputTensorFactory>& outputTensors);
+
+ReturnPathLoweringResult lowerComputeResultReturnPath(spatial::SpatCompute computeOp,
+                                                      mlir::OpResult result,
+                                                      mlir::Value yieldValue,
+                                                      ReturnPathState& state,
+                                                      mlir::IRRewriter& rewriter);
+
+void replaceReturnWithOutputBuffers(mlir::func::ReturnOp returnOp, mlir::IRRewriter& rewriter, ReturnPathState& state);
+
+} // namespace onnx_mlir

src/PIM/Conversion/SpatialToPim/SpatialToPim.td

@@ -9,17 +9,6 @@ include "src/Accelerators/PIM/Dialect/Spatial/Spatial.td"
 include "src/Accelerators/PIM/Dialect/Pim/Pim.td"
 #endif // OP_BASE
 
-def HasSpatialChannelSourceCoreIdAttr: Constraint<
-  CPred<"onnx_mlir::hasSpatialChannelSourceCoreIdAttr($0)">,
-  "spatial channel has precomputed source core id">;
-
-def HasSpatialChannelTargetCoreIdAttr: Constraint<
-  CPred<"onnx_mlir::hasSpatialChannelTargetCoreIdAttr($0)">,
-  "spatial channel has precomputed target core id">;
-
-def createPimReceiveFromSpatialChannelValue: NativeCodeCall<
-  "onnx_mlir::createPimReceiveFromSpatialChannel($_builder, $_loc, $0, $1)">;
-
 def onnxToPimTranspose : Pat<
   (ONNXTransposeOp:$srcOpRes $data, $perms),
   (PimTransposeOp $data, $perms,
@@ -27,17 +16,11 @@ def onnxToPimTranspose : Pat<
 >;
 
 def spatToPimVMM : Pat<
-  (SpatWeightedVMMOp:$srcOpRes $weightIndex, $vector),
+  (SpatVMMOp:$srcOpRes $weightIndex, $vector),
   (PimVMMOp $weightIndex, $vector,
     (NativeCodeCall<"onnx_mlir::getBestOutputTensorFromOperandsOrAllocate($_builder, $0.getDefiningOp())"> $srcOpRes))
 >;
 
-def spatToPimMVM : Pat<
-  (SpatWeightedMVMOp:$srcOpRes $weightIndex, $vector),
-  (PimMVMOp $weightIndex, $vector,
-    (NativeCodeCall<"onnx_mlir::getBestOutputTensorFromOperandsOrAllocate($_builder, $0.getDefiningOp())"> $srcOpRes))
->;
-
 def spatToPimVVAdd : Pat<
   (SpatVAddOp:$srcOpRes $a, $b),
   (PimVVAddOp $a, $b,
@@ -80,18 +63,4 @@ def spatToPimVSoftmax : Pat<
     (NativeCodeCall<"onnx_mlir::getBestOutputTensorFromOperandsOrAllocate($_builder, $0.getDefiningOp())"> $srcOpRes))
 >;
 
-def spatChannelSendToPimSend : Pat<
-  (SpatChannelSendOp $channel, $input),
-  (PimSendOp $input,
-    (NativeCodeCall<"onnx_mlir::getTensorSizeInBytesAttr($_builder, $0)"> $input),
-    (NativeCodeCall<"onnx_mlir::getSpatialChannelTargetCoreIdAttr($_builder, $0)"> $channel)),
-  [(HasSpatialChannelTargetCoreIdAttr $channel)]
->;
-
-def spatChannelReceiveToPimReceive : Pat<
-  (SpatChannelReceiveOp:$srcOpRes $channel),
-  (createPimReceiveFromSpatialChannelValue $srcOpRes, $channel),
-  [(HasSpatialChannelSourceCoreIdAttr $channel)]
->;
-
 #endif // SPATIAL_TO_PIM